Prediction and analysis of model’s parameters of Li-ion ...918015/FULLTEXT01.pdf · Prediction and analysis of model’s parameters of Li-ion battery cells Ali Dareini ... Abbreviations

Prediction and analysis of model’s parameters of Li-ion battery cells

Ali Dareini Blekinge Institute of Technology, Karlskrona, Sweden

Carinthia University of Applied Sciences, Villach, Austria

2016

Master’s Thesis in System Design

Prediction and analysis of model’s parameters of Li-ion battery cells

Thesis for the degree Master of Science

2016

Ali Dareini

AAcademic side:

Department of Applied Signal Processing Blekinge Institute of Technology

Karlskrona, Sweden

Supervisor: Mr. Anders Hultgren Examiner: Mr. Sven Johansson

Department of System Design Carinthia University of Applied Sciences

Villach, Austria

Supervisor: Mr. Stefan Doczy Examiner: Mr. Wolfgang Werth

IIndustry side:

ÅF Automotive Trollhättan, Sweden

Supervisor: Mr. Stefan Ohlsson

National Electric Vehicle Sweden, NEVS Trollhättan, Sweden

Ms. Bodil Ahlström

Contact information:

Ali Dareini

Telephone: +46 700 594 568 E-mail: [email protected]

Anders Hultgren

Blekinge Institute of Technology SE-371 79 Karlskrona, Sweden Telephone: +46 455 385 588

E-mail: [email protected]

Bodil Ahlström

National Electric Vehicle Sweden SE-461 38 Trollhättan, Sweden Telephone: +46 739 665 739


Stefan Doczy

Carinthia University of Applied Sciences A-9524 Villach, Austria

Telephone: +43 664 8825 6313 E-mail: [email protected]

Stefan Ohlsson

ÅF Automotive SE-461 38 Trollhättan, Sweden Telephone: +46 723 700 797


Abstract Lithium-ion batteries are complex systems and making a simulation model of them is

always challenging. A method for producing an accurate model with high capabilities for predicting the behavior of the battery in a time and cost efficient way is desired in this field of work. The aim of this thesis has been to develop a method to be close to the desired method as much as possible, especially in two important aspects, time and cost. The method which is the goal of this thesis should fulfill the below five requirements:

1. Able to produce a generic battery model for different types of lithium-ion batteries 2. No or low cost for the development of the model 3. A time span around one week for obtaining the model 4. Able to predict the most aspects of the battery’s behavior like the voltage, SOC,

temperature and, preferably, simulate the degradation effects, safety and thermal aspects

5. Accuracy with less than 15% error

The start point of this thesis was the study of current methods for cell modeling. Based on their approach, they are divided into three categories, abstract, black box and white box methods. Each of these methods has its own advantages and disadvantages, but none of them are able to fulfill the above requirements.

This thesis presents a method, called “gray box”, which is, partially, a mix of the black and white boxes’ concepts. The gray box method uses values for model’s parameters from different sources. Firstly, some chemical/physical measurements like in the case of the white box method, secondly, some of the physical tests/experiments used in the case of the black box method and thirdly, information provided by cell datasheets, books, papers, journals and scientific databases.

As practical part of this thesis, a prismatic cell, EIG C20 with 20Ah capacity was selected as the sample cell and its electrochemical model was produced with the proposed method. Some of the model’s parameters are measured and some others are estimated. Also, the abilities of AutoLion, a specialized software for lithium-ion battery modeling were used to accelerate the modeling process.

Finally, the physical tests were used as part of the references for calculating the accuracy of the produced model. The results show that the gray box method can produce a model with nearly no cost, in less than one week and with error around 30% for the HPPC tests and, less than this, for the OCV and voltage tests. The proposed method could, largely, fulfill the five mentioned requirements. These results were achieved even without using any physical tests/experimental data for tuning the parameters, which is expected to

reduce the error considerably. These are promising results for the idea of the gray box which is in its nascent stages and needs time to develop and be useful for commercial purposes.

KKeywords: Lithium-ion battery, equivalent circuit model, electrochemical model, AutoLion software, predicting and analysis of parameters, key parameters, tuning the parameter, knowledge-based data, physical and chemical measurements, physical tests, EIG C20

Description of the cover photo:

In the left, there is a SEM image of the NCM chemistry cathode particles and aluminium foil with X430 magnification, taken by an electron microscope in the NEVS lab located in Trollhättan, Sweden. The black area between the coated materials and foil is a gap which is caused by the physical force. The center image shows all the plates/sheets of the opened sample cell. The sample cell of this thesis is a stacked prismatic cell and it is opened for doing some measurements. The right image shows the symbol of electric vehicles at a car parking, equipped with charge stations for electric vehicles in center of Borås city, located in Sweden. The center and right images were taken by the author of this thesis in the year 2015.

Acknowledgment First of all, I want to thank my industrial supervisor, Stefan Ohlsson and the manager

of automotive electrical systems of ÅF Company, Nicklas Karlsson, for trusting me with this challenging project. Special thanks to Stefan Ohlsson which without whom this project would have been impossible. His daily guidance and patience throughout the past months have been invaluable.

I would like to express my special appreciation to my academic supervisors, Anders Hultgren and Stefan Doczy, for all support during this project. Their input and experience have been very helpful during all of the project phases. They always provided constructive feedback as well as positive support, and I consider myself lucky to have had the chance to work with them.

I greatly appreciate the support of ÅF people, Kjell Johansson, Stefan Hellqvist and Xu Wenbo, during my project. A special thanks to Bodil Ahlström and her colleagues at NEVS company, Thomas Mattsson and Kurt Klasson for their great collaborations. They have contributed with their experience and instruments, as well as shown an interest in the project and its results.

Finally, I wish to thank my family and my fiancée for their unconditional love and support throughout this master’s program and this thesis. I am forever grateful.

Ali Dareini,

Sweden, February 2016

Abbreviations and Indices

Abbreviations and Indices

AI Artificial Intelligence ANN Artificial Neural Network BMS Battery Management System BOL Beginning of Life CSDS Commercial Specification Data Sheet CAE Computer Aided Engineering CAEBAT Computer-Aided Engineering for Electric-Drive Vehicle Batteries CT Computerized Tomography CC/CV Constant- Current/Constant- Voltage controlled charge system C/x-rate Current normalized correspond to one nominal discharge capacity

per hour DOD Depth of Discharge EV Electric Vehicle EDV Electric Drive Vehicle ECR Electrical Contact Resistance ECM Equivalent Circuit Model HPPC Hybrid Pulse Power Characterization ICE Internal Combustion Engine LIB Lithium-Ion Battery NiMH Nickel Metal Hydride NiCd Nickel-Cadmium OAN Oil Absorption Number OCV Open Circuit Voltage OCV@100%SOC Open Circuit Voltage at fully SOC PSD Particle Size Distribution SEM Scanning Electron Microscopy SOC State of Charge SVM Support Vector Machine TCB Thermally Coupled Battery

i

Table of Contents

Table of Contents

Table of Contents

1 Introduction ................................................................................. 1

1.1 Background .................................................................................................. 1 1.2 The goal ....................................................................................................... 2 1.3 Scope ............................................................................................................ 3 1.4 Thesis Outline .............................................................................................. 3

2 Problem analysis and method selection ......................................... 5

2.1 Li-ion battery ............................................................................................... 5 2.2 Current methods of battery modeling .......................................................... 9

2.2.1 Abstract method ........................................................................................ 9 2.2.2 Black box method ...................................................................................... 9 2.2.3 White box method ................................................................................... 13

2.3 Method selection ........................................................................................ 14 2.4 Gray box method ....................................................................................... 15 2.5 Software selection ....................................................................................... 17 2.6 AutoLion software ...................................................................................... 19

2.6.1 Equations ................................................................................................. 21

3 Method development .................................................................. 22

3.1 Sample battery cell ..................................................................................... 22

ii

Table of Contents

3.2 Developing the electrochemical model ....................................................... 24 3.2.1 Sources ..................................................................................................... 24

3.2.1.1 Knowledge-based data .............................................................................. 24 3.2.1.1.1 Information in CSDS and MSDS ....................................................... 24 3.2.1.1.2 Information in scientific databases, papers, journals and books ......... 25 3.2.1.1.3 Material knowledge ............................................................................ 26

3.2.1.2 Physical/chemical measurements .............................................................. 26 3.2.1.2.1 Opening the cell ................................................................................. 26 3.2.1.2.2 Simple measurements ......................................................................... 28 3.2.1.2.3 Advanced measurements .................................................................... 30

3.2.1.3 Physical tests (tuning the parameters) ..................................................... 37 3.2.2 Key parameters ........................................................................................ 39

3.2.2.1 Loading ..................................................................................................... 40 3.2.2.1.1 Porosity .............................................................................................. 41 3.2.2.1.2 Porosity of the cathode ...................................................................... 45 3.2.2.1.3 Porosity of the anode ......................................................................... 47 3.2.2.1.4 Porosity of the separator .................................................................... 48

3.2.2.2 Particle size .............................................................................................. 49 3.2.2.3 Open circuit voltage at fully SOC ............................................................ 52 3.2.2.4 Contact resistance of electrodes ................................................................ 52 3.2.2.5 Solid state diffusion coefficient of electrodes ............................................. 53 3.2.2.6 Bruggeman exponents ............................................................................... 54

3.2.3 Analyzing the parameters ........................................................................ 55 3.2.3.1 Sources of the parameters ......................................................................... 55 3.2.3.2 Parameters’ influence over the model ....................................................... 57

3.3 MATLAB generic battery model ............................................................... 58

44 Verification ................................................................................ 61

4.1 Capacity test .............................................................................................. 62 4.1.1 References ................................................................................................ 64

4.1.1.1 CSDS ........................................................................................................ 64 4.1.1.2 Physical tests on the cell ids 32 and 33 .................................................... 65 4.1.1.3 Equivalent circuit model ........................................................................... 65 4.1.1.4 Research report ......................................................................................... 66

4.1.2 Models ...................................................................................................... 66 4.1.2.1 MATLAB generic battery model .............................................................. 67 4.1.2.2 Electrochemical models ............................................................................. 67

4.1.3 Analysis.................................................................................................... 67 4.2 Voltage during the 1C discharge current test ............................................ 69

4.2.1 References ................................................................................................ 69

iii

Table of Contents

4.2.1.1 CSDS ........................................................................................................ 69 4.2.1.2 Physical test on the cell id 32 ................................................................... 70 4.2.1.3 Equivalent circuit model ........................................................................... 71


4.2.3 Analysis.................................................................................................... 74 4.3 Open circuit voltage test ............................................................................ 75

4.3.1 References ................................................................................................ 75 4.3.1.1 Physical test on the cell id 33 ................................................................... 75 4.3.1.2 Equivalent circuit model ........................................................................... 78 4.3.1.3 Research report ......................................................................................... 78


4.3.3 Analysis.................................................................................................... 81 4.4 HPPC test .................................................................................................. 82

4.4.1 HPPC discharge test ................................................................................ 84 4.4.2 HPPC charge test .................................................................................... 86

55 Closure ....................................................................................... 90

5.1 Conclusions ................................................................................................ 90 5.2 Future Work .............................................................................................. 92

Appendix A. Statistical study of information in CSDS & MSDS ....... 97

Appendix B. CAEBAT .................................................................... 99

Appendix C. Safety instruction for opening the cell......................... 101

iv

List of Figures

List of Figures

FFigure 2.1: Specific energy and specific power of rechargeable batteries. ............... 6 Figure 2.2: A schematic view of a LIB [8]. ............................................................. 8 Figure 2.3: Scheme of a black box system. ........................................................... 10 Figure 2.4: State of the art. (a) Thevenin-, (b) impedance-, and (c) runtime-based

electrical battery models [12]. ....................................................................................... 11 Figure 2.5: The comprehensive model proposed by Min Chen [12]. ..................... 12 Figure 2.6: Scheme of a white box system. ........................................................... 13 Figure 2.7: The idea of the gray box method. ...................................................... 16 Figure 2.8: A screenshot of the AutoLion interface. ............................................. 20 Figure 2.9: Simulink block diagram using the AutoLion block (ALST). .............. 20 Figure 2.10: Governing equations, taken from [22]. .............................................. 21 Figure 3.1: Specifications of NCM chemistry. ...................................................... 23 Figure 3.2: The selected sample battery cell. ....................................................... 23 Figure 3.3: The sample cell a few days after opening. .......................................... 27 Figure 3.4: The cathode and the anode plates separated from each other. .......... 27 Figure 3.5: (a): Structure and order of the plates. (b): The sequence of the layers in

prismatic battery cells. ................................................................................................. 28 Figure 3.6: Some of the instruments were used in simple measurements, from the

top: analogue and digital caliper, fine ruler. ................................................................. 29 Figure 3.7: An anode plate with its copper tab. ................................................... 29 Figure 3.8: Scanning electron microscope (JSM-6490LV), used for advanced

measurements. .............................................................................................................. 31 Figure 3.9: Analysis of the cathode materials. ...................................................... 32 Figure 3.10: A SEM image from the edge of a cathode plate. .............................. 35 Figure 3.11: Thickness of an anode plate (cut by scissors). ................................. 35 Figure 3.12: Thickness of two graphite plates, (a): Sample located below its holder,

(b): Sample located above its holder. ........................................................................... 36 Figure 3.13: Two screenshots from the AutoLion video clip, (a): before tuning and

(b): after tuning the parameters’ values. ...................................................................... 38 Figure 3.14: Two different samples from the anode. (a): An anode sample scratched

from the plate. (b): An anode sample from the surface of the plate. ........................... 42

v

List of Figures

FFigure 3.15: Two anode plates. (a): The first plate in the top. (b): The middle plate. ............................................................................................................................. 43

Figure 3.16: Material analysis of the first anode plate. ........................................ 44 Figure 3.17: Material analysis of the anode plate showed in FFigure 3.14(a). ...... 44 Figure 3.18: The cathode material with X1000 magnification. (a): The image from

the microscope. (b): The same image after applying the threshold technique, at level 45. ...................................................................................................................................... 46

Figure 3.19: Anode material with X1000 magnification. (a): the image from the microscope. (b): the same image after applying the threshold technique at level 118. 48

Figure 3.20: The separator layer under the microscope. (a): In vertical position with X2,000 magnification. (b): In horizontal position with X500 magnification. ....... 49

Figure 3.21: A SEM image of the cathode plate. .................................................. 50 Figure 3.22: The particle size distribution curve of a NCM chemistry, from paper

[28] (by varying different content of Ti-doping). .......................................................... 50 Figure 3.23: Particle size of the NCM chemistry taken by the microscope from the

sample cell. ................................................................................................................... 51 Figure 3.24: The influence of the two parameters: OCV@100%SOC and contact

resistance on the results of the voltage test. ................................................................. 53 Figure 3.25: The usage of each source for creating the NCM523 electrochemical

model. ........................................................................................................................... 56 Figure 3.26: The usage of each source for creating the NCM523 electrochemical

model (in detail). .......................................................................................................... 56 Figure 3.27: Parameters’ influence over the NCM523 electrochemical model. ..... 57 Figure 3.28: The Simulink/Simscape file contains the MATLAB model. ............ 59 Figure 4.1: The capacity test for the physical cell id 32. (a): The applied input

current and the voltage response. (b): The temperature of the cell during the test. ... 63 Figure 4.2: Voltage vs. capacity of the cell for different C-rates, taken from the

CSDS of the sample cell. .............................................................................................. 64 Figure 4.3: The ECM discharge curve. ................................................................. 66 Figure 4.4: The histogram of the calculated capacity. ......................................... 68 Figure 4.5: Voltage-SOC curve of the CSDS during the 1C discharge current. ... 69 Figure 4.6: Voltage-SOC curve of the physical test on the cell id 32 during the 1C

discharge current. ......................................................................................................... 70 Figure 4.7: Voltage-SOC curve of the ECM-cell id 35 during the 1C discharge

current. ......................................................................................................................... 71 Figure 4.8: Voltage-SOC curve of the MATLAB model during the 1C discharge

current. ......................................................................................................................... 72

vi

List of Figures

FFigure 4.9: Voltage-SOC curves of the electrochemical models during the 1C discharge current. ......................................................................................................... 73

Figure 4.10: Voltage-SOC curves of all models and references (full SOC range). 74 Figure 4.11: Voltage-SOC curves of all models and references (SOC range 80%-

30%). ............................................................................................................................. 74 Figure 4.12: The dynamic resistance and open circuit voltage test. (a): The applied

current pulses to the physical cell. (b): The voltage response from the physical cell. . 76 Figure 4.13: OCV calculations for the physical test on the cell id 33. ................. 77 Figure 4.14: OCV- SOC curve of the ECM-cell id 35. ......................................... 78 Figure 4.15: OCV-SOC curves from the research report [37]. .............................. 79 Figure 4.16: OCV-SOC curve of the MATLAB model. ........................................ 80 Figure 4.17: OCV-SOC curves of the two electrochemical models. ...................... 80 Figure 4.18: SOC-OCV curves of all models and references (full SOC range). .... 81 Figure 4.19: SOC-OCV curves of all the models and references except of the

MATLAB model (SOC range 30%-80%). ..................................................................... 82 Figure 4.20: The HPPC test profile. (a): Applied current. (b): Calculated SOC. 83 Figure 4.21: All results of the HPPC discharge test (All 10 pulses). ................... 84 Figure 4.22: All results of the HPPC discharge test (Fourth pulse). ................... 85 Figure 4.23: All results of the HPPC charge test (All 10 pulses). ........................ 86 Figure 4.24: All results of the HPPC charge test (4 middle pulses). .................... 87 Figure 4.25: All results of the HPPC charge test (Sixth pulse). .......................... 88 Figure A.1: Chemistry of the studied cells in the statistical study. ..................... 97 Figure C.1: Nine classes of hazardous materials. ................................................. 102 Figure C.2: The plastic cutter was used for opening the cell. ............................. 102 Figure C.3: The ventilation cabins. ..................................................................... 103 Figure C.4: A plastic insulator was used for measuring the weight of the cell with

the balance. .................................................................................................................. 103

vii

List of Tables

List of Tables

TTable 2.1: Comparing rechargeable batteries with different criteria [6]. ................ 7 Table 2.2: Comparison of the four circuit models. ................................................ 12 Table 3.1: Calculating the molecular weight of LiNi0.3333Co0.3333Mn0.3333O2.

...................................................................................................................................... 34 Table 3.2: Calculating the molecular weight of LiNi0.42Co0.18Mn0.4O2. ........... 34 Table 3.3: The results of the thickness measurements. ........................................ 37 Table 3.4: The effect of the cathode’s loading values on capacity and weight of the

designed model. ............................................................................................................. 40 Table 3.5: The measured porosity of the cathode for different magnification rates

and different threshold levels. ....................................................................................... 46 Table 3.6: The measured porosity of the anode for different magnification rates and

different threshold levels. .............................................................................................. 47 Table 3.7: The list of the MATLAB model’s parameters. .................................... 58 Table 3.8: Parameters’ values of the MATLAB model. ....................................... 59 Table 4.1: The capacity test results for the two physical cells 32 and 33. ........... 65 Table 4.2: The capacity test results for the two electrochemical models. ............ 67 Table 4.3: The results of all the capacity tests. .................................................... 68 Table 4.4: Calculated internal resistance (R10s), HPPC discharge test. .............. 86 Table 4.5: Calculated internal resistance (R10s) from the HPPC charge test. .... 88 Table 5.1: Comparison of the black box and the gray box methods. ................... 91 Table A.1: The results of the statistical study. .................................................... 98

viii

Chapter1. Introduction

This chapter gives the background and a brief overview of the thesis goal. The scope and the outlines of the thesis are also covered in this chapter.

1.1 Background

In recent years, the demand for hybrid-electric and fully electric vehicles (EVs) has increased enormously. The automotive industry is undergoing a transformation and moving away from internal combustion engines (ICE) towards electric drivetrains and there are some reasons for it, which [1] enumerates them as:

EEnergy efficiency: EVs convert about 59%–62% of the electrical energy from the grid to power at the wheels. Conventional gasoline vehicles only convert about 17%–21% of the energy stored in gasoline to power at the wheels. Environmentally friendly: EVs emit no tailpipe pollutants, although the power plant producing the electricity may emit them. Electricity from nuclear-, hydro-, solar-, or wind-powered plants causes no air pollutants. Performance benefits: Electric motors provide quiet, smooth operation and stronger acceleration and require less maintenance than ICEs. Reduce energy dependence: Electricity is a domestic energy source.

By 2020, roughly half of the new vehicle sales will likely consist of hybrid-electric, plug-in hybrid, and all-electric models [2].

1 IIntroduction

1


The development of electric cars is significantly more complex than designing conventional cars because they incorporate many different engineering domains into a single system. The complexity of the automobile has increased exponentially in the past decades with higher performance in their components, safety measures, comfort and communications. At the same time, competitive pressures are forcing auto manufacturers to come up with newer designs faster than ever before. This has triggered a design revolution in the automobile industry which stresses detailed modeling and simulation steps prior to committing to metal and plastic. Batteries are the key to this revolution and lithium-ion batteries (LIB) are the best choice for electric vehicles because of their excellent performance, compact, high energy density and high reliability.

The behavior of the battery should be predicted in order to optimize the energy usage and prolong the battery’s life of usage. Therefore access to a reliable simulation model of the battery system is important to let the designers a guide to forecast the behavior of the battery and thus increase the power efficiency of a battery-based system. For example in EV, a battery management system (BMS) with the function of state of charge (SOC) estimation is required in order to let the user know how long the EV can be used before the battery state approaches to empty. Moreover, since the LIB should not be overcharged or over-discharged, an accurate SOC estimation is very important to avoid the system from inadvertent battery abuse and thus ensuring safety and longevity. Having a good simulation model of the battery is essential so that both battery behavior and the physical interaction of the battery with all the other components are properly reflected in the model. Because the battery plays such a vital role in the vehicle, capturing these interactions is essential to designing efficient and effective EVs [3].

1.2 The goal

Researchers have developed a wide variety of simulation models in order to describe the behavior of the LIBs. There are different methods to create these models. One method of creating the model is by using physical and chemical measurements which electrochemical model is an example of this method. Detailed chemical information causes this method the most reliable model but on the other hand complex, time consuming and costly for development. For whom are not cell designer or manufacturers, sometimes this method is inaccessible due to the instruments and extensive chemical measurements which are required to create the model. One of the popular method for creating the battery model is by using physical tests. ECM is a well-known model which can be created by this method. The created model by this method has a good known quality but its costly and time consuming to develop. A method which can produce an accurate model with

2


high capabilities for predicting the behavior of the battery in a time and cost efficient way is desired. Each of the mentioned methods has some advantages and disadvantages but none of them are close to the desired method in all of the aspects.

The goal of this thesis is to create a method to be close to the desired method, as much as possible, especially in two important aspects, time and cost efficiency for developing the model. It means, the method which is the goal of this thesis should be able to produce a generic battery model with low or no costs in a short time (around one week) and applicable to the different anode and cathode chemistries. The model should also be able to predict the most aspects of the battery behaviors like predicting the voltage, SOC, temperature and preferably simulate the degradation effects, safety and thermal aspects of the battery. Regarding the accuracy, around ±15% error is tolerable to reduce the cost and time for producing the model. To be more clear the ±15% error refers to the prediction of voltage and SOC of the battery in different tests like open circuit voltage (OCV), voltage of the cell during 1-C discharge current and standard hybrid pulse power characterization (HPPC). This is applied to all of the “error” terms which are used in this report.

1.3 Scope

Defining the scope of the thesis project is necessary for insuring the success of the project. An important point which needs to be reformulated is, the desired method should be capable to create a model to predict preferably all aspects of the battery behavior, but this is out of the time of this thesis. Therefore the thesis is limited and then modeling a single battery cell is on the focus of this thesis and not a complete battery system or BMS part. Also the temperature needs to be considered as one of the influence factors, but the model and simulation/physical tests are done at room temperature. When the results are good enough, the effect of the temperature can be considered on the next steps. This situation is the same for degradation effects and then the cell is considered in the beginning of life (BOL). Safety simulation and 3D-Thermal modeling are also not considered in this step of the work.

1.4 Thesis Outline

The report is organized as follows:

In CChapter 1 the background and a brief overview of the thesis goal are given. This chapter also covers the scope and outlines of the thesis.

3


In CChapter 2 the problem which this thesis is going to solve is explained and analyzed extensively. For understanding the problem a brief description of LIBs is given. The already available methods for modeling the LIBs are studied and based on all of the accessible and available information, the problem is attacked. Finally the best suitable method which can solve the problem is proposed. In addition the tools and software which are used in this thesis are discussed and selected.

In CChapter 3 the proposed method (gray box) is employed to create the electrochemical model. Three groups of sources are presented to extract the model’s parameter’s values. In each group, the possible parameter which can be extracted are explained and shown by some examples. After finding the value of all the parameters, they are analyzed from different perspectives. Beside the electrochemical model, a model from MATLAB which is based completely on the information in commercial specification data sheet (CSDS) is also created.

In CChapter 4, the results of the designed models are compared with different references for four tests:

Capacity Voltage of the cell during 1-C discharge current OCV HPPC

These tests will help us to understand how much the accuracy of the proposed method is close to the desired method.

In CChapter 5, the conclusions are presented and potential future works are introduced.

This report also contains three appendices. AAppendix A contains the results of the statistical study of “available information in CSDSs and MSDSs”. AAppendix B explains the CAEBAT project and AAppendix C shows the safety instruction which are used for opening the physical cell.

4

Chapter2. Problem analysis and method selection

In this chapter, the problem which this thesis is going to solve, is explained and analyzed extensively. For understanding the problem a brief description of LIBs is given. The already available methods for modeling the LIBs are studied and based on all of the accessible and available information, the problem is attacked. Finally the best suitable method which can solve the problem is proposed. In addition, the tools and software which are used in this thesis are discussed and selected.

2.1 Li-ion battery

A battery is defined as an electrochemical storage device that converts the chemical energy contained in its active materials directly into electric energy by means of an electrochemical redox reaction. This reaction involves the transfer of electrons from one material to another through an electric circuit. Scientifically batteries are referred to as electrochemical or galvanic cells, due to the fact that they store electrical energy in the form of chemical energy and because the electrochemical reactions that take place are also termed galvanic [4].

The use of rechargeable batteries in consumer products, business applications and industrial systems continues to grow substantially. The global market for all batteries was reached almost $74 billion in 2015, and rechargeable batteries will account for nearly 82% of that, or $60 billion, according to market researcher Frost & Sullivan [5].

22 PProblem analysis and method sselection

5


LIBs of various types have been moving rapidly toward commercialization ascribing to their potential advantages in power density, cost, safety, performance and design flexibility. Recently, more and more attention has been paid to EV and hybrid electric vehicle (HEV). LIB shows all the signs of the beginning of a new product cycle, with sales growing exponentially owing to the better capacity compared to any other existing commercial battery system.

Batteries advance on two fronts, specific energy for longer runtimes and specific power for good power delivery. FFigure 2.1 illustrates the energy and power densities of lead acid, nickel cadmium (NiCd), nickel metal hydride (NiMH) and the li-ion family.

Figure 22.11: Specific energy and specific power of rechargeable batteries.

Specific energy, also known as gravimetric energy density is the capacity a battery can hold in watt-hours per kilogram (Wh/kg) and specific power, also known as gravimetric power density is the battery’s ability to deliver power in watts per kilogram (W/kg). Figure 2.1 shows, among different type of rechargeable batteries, li-ion family batteries have higher specific energy and higher specific power that makes them suitable for different applications from portable electronic devices to power system of aircrafts, spacecraft and EVs. Beside the advantage of LIBs in high specific energy and power, they are preferable to the other rechargeable batteries in other criteria which are shown in Table 2.1.

6


TTable 22..11:: Comparing rechargeable batteries with different criteria [6].

NiCd NiMH Lithium-ion Lithium-ion

polymer

Operating voltage (V) 1.2 1.2 3.7 3.7

Energy density (Wh/kg) 60 80 >100 >100

High current performance Good Good OK Good

Cycle life (deep discharge) <500 <300 >500 >1000

Quick charge (hour) 1 1.5-3 <1.5 <1.5

Self-discharge (% per month) 5 15-30 <5 <5

Specific advantage Low cost Higher energy High energy and power density

In TTable 2.1, the term “polymer” refers to LIBs in pouch format. According to [7], the designation “lithium polymer” can be interpreted in two ways. Originally, it represented a technology using a polymer electrolyte instead of the more common liquid electrolyte. The second meaning appeared after some manufacturers applied the “polymer” designation to lithium-ion cells contained in a non-rigid pouch format. This is currently the most popular use, in which “polymer” refers more to a “polymer casing” (that is, the soft, external container) rather than a “polymer electrolyte”.

A LIB basically composed of anode, cathode, electrolyte with lithium salt and separator. The anode, also called the negative electrode releases electrons into the external circuit during the discharge process, which is associated with oxidative chemical reactions. The cathode, also called the positive electrode, gains electrons from the external circuit during the discharge process, which is associated with reductive chemical reactions. An electrolyte is a material that acts as a charge carrier to provide pure ionic conductivity between the anode and cathode in a cell. A separator is a physical barrier between the positive and negative electrodes to prevent an electrical short circuit. The separator can be a gelled electrolyte, or a micro porous plastic film or other porous inert material filled with electrolyte. Separators must be permeable to the ions and inert in the battery environment. The current collectors are copper at negative terminal and aluminium at positive terminal. A LIB cell is shown schematically in Figure 2.2.

7


FFigure 22..22:: A schematic view of a LIB [8].

FFigure 2.2 shows the charge and discharge processes, main materials, namely, the anode (e.g. graphite), cathode (e.g. . . . (NCA)), current collectors (copper and aluminium), an organic electrolyte and a separator which separates the anode and cathode.

Under normal operation, charging the battery causes lithium ions in the electrolyte solution to migrate from the cathode through a micrometer-thin porous polymer separator and insert themselves (intercalate) in the anode. Common cathodes are based on (LCO), (LMO), (LFP), . . . (NMC), . . . (NCA) and related oxides. The anode is generally a form of graphite. Charge-balancing electrons also move to the anode but travel through an external circuit in the charger. On discharge, meaning when the battery is used to provide power, the reverse process occurs, and electrons flow through the device being energized. For additional information on general working principles of a LIB, the readers are referred to reference [9].

Discharge

Charge

8


2.2 Current methods of battery modeling

In this section, an overview of current methods for battery modeling is given. Researchers around the world used different methods to develop a wide variety of models with varying degrees of complexity. They capture battery behavior for specific purposes, from battery design and performance estimation to circuit simulation.

In literature, many battery models can be found. Different approaches have been used to model the battery properties, varying from very detailed electrochemical models to high level stochastic models. In this report an overview of most well-known of different battery models is given.

Reference [10] divided the battery models into two categories, mathematical models and ECMs. Here mathematical models contain electrochemical and empirical models. Reference [11] divided them into four categories, empirical models, electrochemical models, electrical-circuit models and abstract models using artificial intelligence (AI).

Researchers have used several techniques on battery modeling. References [12], [13], and [14] provide a good review on several battery modeling techniques. In this thesis, the battery models categorized based on their approaching methods and divided into three categories, abstract, black box and white box. In continue, these three methods are discussed in detail.

2.2.1 Abstract method

This method of modeling is easy to configure but the results are the least accurate. They describe the battery at a higher level of abstraction and with reduced order of equations. The major properties of the battery are modeled using only a few equations. Mostly they cannot offer any current or voltage information that is important to circuit simulation and optimization. Empirical and analytical models are some examples of abstract models which they are more applicable to the imprecise capacity evaluation. Most abstract models only work for specific applications and provide inaccurate results. For some of them, this error range can be even higher, for example, the maximum error of Peukert’s law predicting runtime can be more than 100% for time-variant loads [15].

2.2.2 Black box method

A “black box” here is a system which can be viewed in terms of its inputs and outputs, or transfer characteristics, without any knowledge of its internal workings. The scheme of a black box system is shown in FFigure 2.3. The understanding of a black box system is based on the “explanatory principle”, the hypothesis of a causal relation between the input and the output. The black box method for a battery comes from the

9


observations on battery voltage, current and temperature behavior. It means without necessary knowing about the material and internal structure of the battery and just by looking to the input and output of the battery the model is achievable.

FFigure 22..33:: Scheme of a black box system.

The well-known model of this method is ECM. This kind of model consists of electric circuit elements, such as voltage sources, resistors and capacitors for co-design and co-simulation with other electrical circuits and systems. For electrical engineers, electrical models are more intuitive, useful, and easy to handle, especially when they can be used in circuit simulators and alongside application circuits. This model is good for circuit simulation.

There are many electrical models of batteries, from lead-acid to LIBs. Most of these electrical models fall under three basic categories, Thevenin, impedance and runtime-based models, shown in FFigure 2.4.

(a) (b)

10


(c)

FFigure 22..44:: State of the art. (a) Thevenin-, (b) impedance-, and (c) runtime-based electrical battery models [12].

Each of the electrical models in FFigure 2.4 has its own advantages and disadvantages. Thevenin-based electrical model is the most basic form, shown in Figure 2.4(a) and uses a series resistor, and an RC parallel network, and to predict battery response to transient load events at a particular SOC. The main disadvantage of this model is, none of the Thevenin-based models can predict the battery runtime simply and accurately in circuit simulators. Impedance-based models, shown in Figure 2.4(b), employ the method of electrochemical impedance spectroscopy to obtain an AC-equivalent impedance model in the frequency domain, and then use a complex equivalent network (Zac) to fit the impedance spectra. The fitting process is difficult, complex, and non-intuitive. In addition, impedance-based models only work for a fixed SOC and temperature setting and therefore they cannot predict DC response or battery runtime. Runtime-based models, shown in Figure 2.4(c), use a complex circuit network to simulate battery runtime and DC voltage response for a constant discharge current. They can predict neither runtime nor the voltage response for varying load currents accurately. A brief comparison illustrated in Table 2.2 indicates that none of these models can be implemented in circuit simulators to predict both the battery runtime and current–voltage performance accurately [12].

11


TTable 22..22:: Comparison of the four circuit models.

Thevenin-

based model Impedance-based model

Runtime-based model

Comprehensive model

DC NO NO YES YES

AC Limited YES NO YES

Transient YES Limited Limited YES

Battery runtime

NO NO YES YES

Therefore, a comprehensive circuit model combining the transient capabilities of Thevenin-based models, AC futures of impedance-based models, and runtime information of runtime-based models is proposed by Min Chen [12] and depicted in FFigure 2.5.

Figure 22.55: The comprehensive model proposed by Min Chen [12].

The proposed model by Min Chen showed in Figure 2.5 is a blend of three previous models in Figure 2.4 and in whose unique combination of components and dependencies ease the extraction procedure, makes a fully Cadence-compatible model possible, and simultaneously predicts runtime, steady state and transient response accurately and “on the fly” capturing all the dynamic electrical characteristics of batteries, usable capacity ( ), open-circuit voltage, and transient response (RC network).

Theoretically, all the parameters in ECM are multivariable functions of SOC, current and temperature. For calculating the model parameters, normally a lot of experimental data is required which as mentioned before is time consuming and costly.

Black box models, with accuracy around 1%–5% error had been used for battery modeling by researchers in [16], [10], [3], [12]. Also for more details about the ECM, the readers are referred to the reference [17].

12


2.2.3 White box method

The opposite of a black box system is a system where the inner components are available for inspection. The scheme of a white box system is shown in FFigure 2.6. In white box system unlike the black box, the internal structure and process for the relation between the inputs and the outputs are known. Then in white box method, knowing the internal structure and the relation between the inputs and the outputs are necessary. This can be achieved by physical/chemical measurements.

Figure 22.66: Scheme of a white box system.

The most well-known model of the white box method is the electrochemical model. This type of the modeling is good for battery design. The electrochemical models are based on the chemical processes that take place in the battery and they can capture the characteristics of cells using mathematics based on the electrochemical theory. This method is a physics-based approach for modeling a battery.

Most of the current rigorous electrochemical models are derived from the porous electrode and concentrated solution theories proposed by Newman and Tiedemann [18] and Doyle et al [19] which mathematically describe charge/discharge and species transport in the solid and electrolyte phases. This type of the models mostly consists of six coupled, non-linear differential equations. Then the models describe the battery processes in great details and the user has to set over 50 battery related parameters, e.g., the thickness of the electrodes, the initial salt concentration in the electrolyte and particle size of the electrodes. Solving these equations gives the voltage and current as functions of time, and the potentials in the electrolyte and electrode phases, salt concentration, reaction rate and current density in the electrolyte as functions of time and position in the cell. From the output data, it is possible to obtain also the battery lifetime. This makes this kind of model the most accurate one yet. The electrochemical models are often used as a comparison against other models, instead of using experimental results to check the accuracy.

On the other hand, the highly detailed description makes this models complex and difficult to configure. To be able to set all these parameters one needs a very detailed knowledge

13


of the battery that is to be modelled. However, because they involve a system of coupled time-variant spatial partial differential equations a solution for which requires high computational efforts. Normally 250 DAEs is the minimum number of equations required for a converged finite difference solution of the full-order model [20].

2.3 Method selection

The current methods for battery modeling are reviewed in section 2.2 and it will be shown why these methods cannot fulfill the requirements of the thesis goal, showed in section 1.2, and why proposing a new method is inevitable.

AAbstract method: Some of the models by this method are not generic and may not be extended to other batteries without additional training data like AI-based learning approach includes artificial neural network (ANN) modelling as well as support vector machine (SVM) [21]. Most of them also do not consider the temperature effect which is very important for an accurate model. Moreover the error of some of them specially for full range of operation condition can be very high, like Peukert’s law predicting runtime which can be more than 100% for time-variant loads [15].

Black box method: The big disadvantage of this method is, every new model created by this method needs to be experimentally characterized over a wide range of operating conditions to create a map or look up table for resistance and capacitance dependence on SOC, temperature, and C-rate. This approach can be inflexible, costly and time consuming, particularly when seeking to demonstrate battery cycle life. In addition, the models are incapable of accounting for battery internal behavior on battery control or monitoring, which largely dictates battery life and safety rendering. Also ECMs are ineffective in developing safety and life conscious system operation and control.

White box method: This method has the possibility to create the most accurate model. Similar to the black box method which depends on physical tests, the white box method depends on physical/chemical measurements which is again costly and time consuming. Also sometimes, some of the detailed physical/chemical information cannot be measured precisely after assembling the cell and they are a property of the cell designers/manufacturers, even sometimes, some of the cell manufacturers do not know or necessary measure some parameters. Moreover, the computational efforts of electrochemical models are high due to their nature which are consists of coupled time-variant spatial partial differential equations but this problem is tolerable in the case of using the reduced models. For example in [20] the model reformulated and the number of equations are reduced from 250 DAEs to 47 DAEs.

14


Different modeling methods and their disadvantages are discussed. Each of them has some notable weak points. These weak points, for the abstract method are limited capabilities and high error for full range of operation condition, and for the black and white box methods are time consuming and costly development process of modeling. By considering the advantages and disadvantage of the current methods, an idea arose and called “gray box”. In the next section, this idea and the explanation of this method are given in detail.

2.4 Gray box method

The black box method as described in section 2.2.2, is based on intensive physical tests. In this method, physical/chemical measurements are not used for creating a model. Vice versa, the white box method creates a model, based on extensive physical/chemical measurements without giving a role to physical tests. It means, in each of these two methods, some parts of the information are ignored and the potential of these sources is not utilized. Then the idea of gray box method is to use both the sources of information but not intensively or extensively. Then in this method, some parts of physical/chemical measurements and some parts of physical tests are used to model the battery and that is the reason this method is named “gray box”. This method also uses another source called “knowledge-based data”. This source covers the information which are available in CSDS, MSDS, books, papers, journals and other knowledge-based data which are mostly free. The idea of gray box method is depicted in FFigure 2.7. In this figure, the sources of data for creating the models are shown.

15


FFigure 22..77:: The idea of the gray box method.

The black box method by using the physical tests can create an ECM and the white box method by using the physical/chemical measurements can create an electrochemical model. The concept of the gray box method, which is utilizing the potential of all sources, can be used to create different models. As the first try, the electrochemical model is selected to be created by this method. The reason is, higher number of parameters in this model than the other models, gives us more options to measure, predict or tune.

Both the gray box and the white box methods create an electrochemical model but in different ways. White box methods measure the whole parameters and can then create an accurate model, the gray box method measures some of the parameters and the rest of the parameters are estimated or tuned. In section 3.2.3.2, it is shown that for the created model in this thesis 74% of the parameters are estimated and the rest of the parameters are measured. This is the main advantage of the gray box method which can reduce the number of measurements/tests and consequently save time and money.

The parameters of the electrochemical model created by the gray box method, as it is shown in FFigure 2.7, can be extracted by looking into three groups of sources:

1. KKnowledge-based data including: Information in CSDS and MSDS Information in scientific databases, papers, journals and books Material knowledge Material database in specialized battery software

16


2. PPhysical/chemical measurements including: Simple measurements Advanced measurements

3. PPhysical tests including: Formation test (cycling the cell with standard charge and discharge C-rate) Standard HPPC test

The parameter’s value can be found with different levels of accuracy may in one, two or all of the three sources. This is the advantage of the gray box method which can give this option to the users to select between the sources of data at least for some of the parameters. In chapter 3, the parameters and the way for finding their values are discussed in detail.

There are some specialized software for modeling the LIB. In this thesis they are studied and one of them is selected to be used. This software accelerates the modeling process. In the next section, this topic will be discussed.

2.5 Software selection

LIBs have attracted a lot of attentions from software developers recently. Alongside the simulators like MATLAB, Mathematica or Modelica which solve the electrochemical equations to simulate a LIB, there are a considerable number of specialized software for just modeling LIBs. The specialized battery software makes the modeling process much faster and convenient for the users. Seven specialized battery software were investigated in this thesis. They are listed below.

1. ANSYS (Battery module) 2. AutoLion 3. Battery design studio 4. CAE-Bat from NREL 5. Comsol Multiphysics (Battery module) 6. Maplesoft (Battery module) 7. Thermoanalytics (Battery module)

All of the listed software were investigated for this thesis and an elimination method was employed for selecting the most suitable one for this thesis.

The first criteria is the capability of the software for modeling the electrochemical model. All of the 7 software, listed above have this possibility and some of them also offer ECM.

The second criteria is possessing material database. The intended material database should contain the materials which are used in available commercial physical cells and

17


could offer the parameter’s value for each selected material. For example the particle size and density of electrodes are different for each chemistry and these should be obtained based on a sufficient number of measurements on different chemistries. Between the 7 software, three of them have material database, AutoLion, Comsol Multiphysics and Maplesoft. The capability of these three software and their material database were tested by using their trial versions.

Material database of Maplesoft offers some default values for the parameters but these values are constant for different chemistries and are just some examples of a cell with LiCoO2 and LiC6 system published in [20] and is obviously not extensible for the other chemistries. Its material database could not fulfill the requirements of the intended material database for this thesis and hence Maplesoft is not selected as a suitable battery software for this thesis.

Batteries and fuel cells module of Comsol Multiphysics seemed to be a powerful software for electrochemical modeling of LIBs. Its material database contains most of the available commercial chemistries like lithium cobalt oxide (LCO), lithium manganese oxide (LMO), lithium iron phosphate (LFP), lithium nickel manganese cobalt oxide (NCM111), lithium nickel cobalt aluminium oxide (NCA) and lithium nickel oxide (LiNiO2) for cathode, and lithium titinate (LTO), hard carbon, silicon electrode (LixSi) and graphite for anode and 5 different solutions for lithium hexafluorophosphate (LiPF6) electrolyte. This software gives a lot of freedom to the users to design a cell and consequently this makes the modeling procedure more complicated. The evaluation process of this software could not be finished completely and it was due to a high complexity level for creating the model for us as electrical engineers and hence this software is not selected for this thesis work.

AutoLion unlike Comsol, which is a simulator software with a module for batteries and fuel cells, is designed specifically for electrochemical modeling of LIBs. Its material database, similar to Comsol, contains most of the available commercial chemistries. According to its user’s manual, extensive testing of the commercially-relevant materials is performed to achieve accurate descriptions of the required material properties over a wide range of conditions. Also a suite of state-of-the-art diagnostic techniques, including galvanostatic intermittent titration technique (GITT), potentiostatic intermittent titration technique (PITT), and 3-electrode electrochemical impedance spectroscopy (EIS), among others, have been utilized to measure material-specific properties. The interface of this software is also convenient and easy to use by an electrical engineer so far modeling a cell is doable just in a few minutes.

Both Comsol Multiphysics and AutoLion are powerful software for LIB modeling. In some aspects the capability of Comsol is even more than the AutoLion. For example, in Comsol any geometry is designable for a cell, but in AutoLion the three known geometries,

18


cylindrical, stacked prismatic and rolled prismatic are predefined and users can select just one of them. Overall, due to the user-fliendly interface, AutoLion was selected as the most suitable software for this thesis work. This software and its capabilities for modeling LIBs are explained in the next section.

It is interesting to know that the only free and open source software between the studied ones, is CAE-bat from NREL. CAE-bat is a project for accelerating the development and lowering the cost of lithium-ion batteries, which was initiated by the U.S. department of energy's office of energy efficiency and renewable energy. The explanation of this software is given in Appendix B.

2.6 AutoLion software

AutoLion is developed by EC Power located in Pennsylvania State, USA. As mentioned before this software is specialized for electrochemical modeling. This software is based on the thermally coupled battery (TCB) modeling approach that allows physics-based fast and dynamic battery module/pack modeling for system-level simulation. AutoLion can be used to predict capacity loss, voltage decay, and individual rates of anode and cathode degradation all for user-specified and wide-ranging charge/discharge rates, temperatures, cycling depth of discharge (DOD), and time. Also this software is capable of capturing dynamic response of a battery with change in ambient and cell-internal temperature during operation as well the impact of dynamic system operation on cell aging. In some versions of this software, 3D thermal modeling and safety simulation is also possible.

The used version for this thesis is called AutoLion-ST which simulates electrochemical models in MATLAB/Simulink environment. This version produces an “.xml” file which contains the value of the parameters, and then this file can be read by a “.mexw64” file which contains the equations and relation between the parameters. This file can be used as a block in Simulink and then other customized blocks like inputs and outputs can be connected to the battery model. It is worth mentioning that the “.mexw64” file and its codes are not visible for the users.

A screenshot of the AutoLion interface is shown in FFigure 2.8. Also Figure 2.9 shows the Simulink block diagram containing the AutoLion block and customized inputs and outputs.

19


FFigure 22..88:: A screenshot of the AutoLion interface.

FFigure 22..99:: Simulink block diagram using the AutoLion block (ALST).

20


The equations of the electrochemical model which are used in the software, are explained in the next section.

2.6.1 Equations

The governing equations are given in FFigure 2.10. These five equations used in the software are well-known and available in most of the books and papers about electrochemical models and as an example, equations shown in Figure 2.10 are taken from the book [22].

Figure 22.110: Governing equations, taken from [22].

The TCB modeling used in the software, has its roots in the isothermal model of Doyle and Newman [19], and substantial extensions through the electrochemically and thermal couplings by Gu and Wang [23], Srinivasan and Wang [24], and Smith and Wang [25]. One major benefit of the software’s model as opposed to most Pseudo- 2D or Newman-type models is that, the software utilizes a fully coupled electrochemical and thermal method.

In the software, there is no simplified treatment e.g. linearized analytical or quasi-analytical solution of solid diffusion in the active material particles. Also despite of any simplification in the equations, the software has good computational speed, e.g. taking 10-15 seconds computing for a standard 1C discharge.

21

Chapter3. Method development

In this chapter the proposed method is employed to create the electrochemical model. Three groups of sources are presented to extract the model’s parameter’s values. In each group, the possible parameter which can be extracted are explained and shown by some examples. After finding the value of all the parameters, the parameters are analyzed from different perspectives. Beside the electrochemical model, a model from MATLAB which is based completely on the information in CSDS is also created.

3.1 Sample battery cell

One of the most important tasks before starting the development process, besides preparing the software and laboratory, is selecting the sample battery cell. For choosing the sample cell two criteria were considered. Firstly, running the physical tests is a costly and time consuming task, so the decision of using the previous physical tests performed in the company was taken. Secondly, the chemistry of the cell should be suitable for electric vehicles.

In [26] different cathode chemistries are compared together and it is showed why NCM chemistry is suitable for electric vehicles. The specification of NCM chemistry is shown in FFigure 3.1.

3 MMethod development

22


FFigure 33..11:: Specifications of NCM chemistry.

Based on the two mentioned points, the cell “ePLB C020” was selected as the sample battery cell for this thesis. This cell fabricated by EiG corporation (South Korea) and uses NCM and graphite as active materials for the cathode and, respectively, anode. The sample cell has a high specific energy which makes it suitable for electric vehicle applications. The gravimetric and volumetric energy density of the cell sample is approximately 186 Wh/kg and 367 Wh/L respectively. The weight of the battery cell is approximately 412 grams and the cell has a prismatic geometry with 20Ah nominal capacity. FFigure 3.2 shows the sample battery cell.

Figure 33.22: The selected sample battery cell.

The topic of NCM chemistry and its different types are explained in section 3.2.1.2.3. The physical tests run on this cell in year 2009. The newer and improved version of this cell type is still in production line of EiG Corporation. In the following section, the electrochemical model’s parameters of the selected sample cell are determined.

23


3.2 Developing the electrochemical model

The number of electrochemical model’s parameters in the software for the prismatic geometry are 138. These parameters are divided into two groups, design and simulator with 71 and 67 parameters respectively. The design tab contains the inner and outer dimensions of the cell and the details regarding the electrode, separator plates and electrolyte, while the simulator tab incorporates the initial conditions, operating conditions, Bruggeman exponents, thermal model, degradation rate, etc.

In this section, the gray box method is used to extract the values of the parameters from three mentioned sources of data. Afterwards, the parameters are analyzed from different perspectives.

3.2.1 Sources

As mentioned in section 2.4, the sources of data for building the model are divided into three groups. This section offers more details about the types of data and, also some examples of extracted parameters.

3.2.1.1 Knowledge-based data

Considerable efforts are done to reveal the maximum potential of this source. In section 3.2.3.1, it is shown that this source has the biggest impact (55%) for building the model and this constitutes one of the advantages of the gray box method, because it is giving a big role to the knowledge-based data. This source is explained in the next 3 subsections.

3.2.1.1.1 Information in CSDS and MSDS

The published information in a cell’s datasheets (CSDS and MSDS) can be different for different cell brands and cell types. Then to have a clear view about the available information in CSDS and MSDS, a statistical study was performed in this thesis. The results of this study are shown in Appendix A and some of the information which are used for modeling are explained in the following.

The geometry and outer dimensions of the cell are the first parameters that are needed for the modeling of the cell and they can be determined using the data in CSDS. The weight of the cell is also used in the model, but indirectly. In the software, weight of the cell and its components are calculated by the entered parameters from the design tab. Then the weight of the cell can be used to check, if the design process is done correctly or not. Other information in CSDS, like suggested charge and discharge conditions and

24


also maximum applicable current/voltage can be used in testing process for model validation.

One of the most useful parameter in CSDS is the discharge characteristic's graph which is available in more than 80% of CSDSs, according to the statistical study showed in Appendix A. The discharge characteristic for the sample cell is shown in FFigure 4.2. The capacity of the cell for different C-rates can be used to check if the model designed correctly or not. Also 1-C discharge curve, which is a part of discharge characteristic’s graph, can be used for tuning the key parameters. In section 3.2.1.3 more explanation are given about the tuning and the way the gray box uses it.

In MSDSs, the material components of the cell can be found. In more than 90% of MSDSs, the cathode and the anode chemistries are given. These information are vital for the electrochemical model. Also in more than 60% of MSDSs, the approximated weight percentage of the cell materials are given. However, in some cases which the approximated weight percentages are given in a big range (e.g. 20%-50%), the information are not useful anymore.

3.2.1.1.2 Information in scientific databases, papers, journals and books

Considerable information can be found in knowledge-based sources like scientific databases, journals, papers and books. Physical tests are a part of these information. In this thesis, a research report containing some physical tests on the same type of the sample cell was found and used as one of the references on the verification process in section 4.1.1.4.

Alongside the physical tests, which can be used in parameter’s tuning or verification process, some of the information can be used directly inside the model. At least 7 papers containing some physical/chemical measurements on the same type of the sample cell were found. For example in paper [26], the internal structure, inner dimensions, thickness of the cell plates and number of the plates are given. Some of these information can be used directly in the model and some of them can be used to check if the model was designed correctly or not. For example, the number of plates are calculated and reported by the software based on the inputted thickness of the plates.

First charge/discharge capacity of the cathode and also density of the electrodes/foils were another part of the information which found and used directly in the model. In conclusion, around ten parameters of the model in this thesis determined using the information available in the research papers.

25


3.2.1.1.3 Material knowledge

Most of the books, which are treating the topic of LIB materials, offer a lot of practical information on subject. The book [22] is one of the most important references studied in this thesis. In page 213 of this book, separator thickness and separator porosity, two parameters of the model, are presented in detail and it is mentioned that “typically, lithium ion battery separators have a porosity of 40%”. This value is the same as the default value in the software and verified by the website of Celgard, one of the major separator manufacturer. In section 3.2.2.1.1, the porosity topic is discussed in detail and it is explained why the porosity of separator cannot be seen by the used microscope. Measuring the separator porosity is a costly task, so in this thesis, an estimation from this book is used. This shows the essence of the gray box method to replace the estimated values from the free knowledge based data with the costly measurements. Besides the separator thickness and porosity, this book gives an estimation for the below parameters which were used directly in the model:

Particle size of electrodes Electrolyte ionic/diffusional conductivity Bruggeman exponents Diffusion coefficient of electrodes Electron conductivity of electrodes/foils

3.2.1.2 Physical/chemical measurements

The physical/chemical measurements are the source of data for the white box method. Gray box method also used it but just partially. This source of data, based on the instruments’ type, is divided into two categories: simple and advanced. Except the outer dimensions and the weight of the cell, for measuring the other parameters, the cell must be opened. Before explaining the measurements' categories, the steps needed for opening the cell are presented.

3.2.1.2.1 Opening the cell

Opening the cell presents explosion and fire hazard due to the flammable materials used in the electrolyte. The safety instructions must be taken into consideration before opening the sample cell. Appendix C contains the safety instruction available in MSDS and, also, the laboratory regulations. Before taking the measurements, the electrolyte was allowed to evaporate completely. FFigure 3.3 shows the sample cell, a few days after opening.

26


FFigure 33..33:: The sample cell a few days after opening.

The first interesting point which can be observed in FFigure 3.3 is that the dimensions of the electrodes and separator plates are not the same. The separator with 195mm has the maximum height and the cathode the minimum height with 187mm. The anode is placed in the middle and it has a height of 193mm. This situation is similar for width, too. The first plate of the cell is the separator but in Figure 3.3, this plate had already been removed, so the first plate showed in this figure, is the anode.

After opening the cell, the plates are separated from each other and all of them are placed on a table in the laboratory. Figure 3.4 shows all of the electrode plates, 17 for the cathode and 18 for the anode. Normally, in prismatic battery cells, the number of anode plates has one more unit comparing to the number of cathode plates.

Figure 33.44: The cathode and the anode plates separated from each other.

Cathode Anode

27


All of the plates are covered with separator plates. The total number of separator plates is 36. The structure of prismatic battery cells, showed in FFigure 3.5 can help in understanding the layout and order of the plates.

(a)

(b)

Figure 33.55: (a): Structure and order of the plates. (b): The sequence of the

layers in prismatic battery cells.

Some of the plates showed in Figure 3.4, were selected randomly for doing the measurements. Next two sections will present the types of the measurements performed on the selected plates.

3.2.1.2.2 Simple measurements

The term “simple” refers to the type of the instruments used during the measurements. Fine ruler, weight scale and micrometer caliper for measuring the thickness are the instruments used for simple measurements. Some of them are shown in Figure 3.6. Weight of the cell and its components, outer dimensions of the cell, inner dimensions of the plates and number of the cathode, anode and separator plates are some of the information which can be measured by a ruler and a weight scale. The above mentioned measurements cover around 20 parameters of the model.

28


FFigure 33..66:: Some of the instruments were used in simple measurements, from the

top: analogue and digital caliper, fine ruler.

Measuring the thickness is not an easy task like the other simple measurements and it can be a challenge, especially for the separator and foils, which have less than 100µm thickness. As the foils are coated with the electrode materials, measuring their thickness with the simple instruments is impossible, but there is a tricky solution which can be employed and that is measuring the part of the foil which is connected to the current collector tabs and it is not coated with the electrode materials. This part is shown in FFigure 3.7 in a red circle.

Figure 33.77: An anode plate with its copper tab.

29


This solution can be useful for measuring the thickness of the foils but it has to be considered that maybe the foils' thickness in the tab area is different from the other areas. Then for measuring the thickness of the foils in the other area, the simple instruments cannot be used. Because the foils are covered in coated materials, using simple instruments for measuring their thickness may result in scratching their surfaces or leaving residual particles behind, so the obtained results will not be correct. In section 3.2.1.2.3, it is shown, how the thickness of the plates/layers can be measured by microscope.

The density of the separator/foils, also can be measured by the simple measurements. The density can be measured using specialized instruments or it can be calculated by dividing the weight of the sample (g) by its volume (cm3).

Briefly, the below items can be calculate by simple instruments:

Inner dimensions of the cathode, anode and separator plates Thickness of the cathode, anode and separator plates Thickness of the aluminium and copper foils Weight of the cell and enclosure Weigh of the cathode, anode and separator plates Number of the cathode, anode and separator plates Density of the aluminium and copper foils Density of the separator

More than 70 measurements are necessary for calculating the above items. The reason behind the high number of measurements is that the thickness of the foils and plates has to be measured in different points, near the edges or in the center. All of the simple measurements in this thesis are done just in a few hours. This type of measurement can extract more than 25 parameters for the electrochemical model with good enough precision.

3.2.1.2.3 Advanced measurements

For measuring some of the parameters for the electrochemical model, advanced instruments are required. A scanning electron microscope shown in FFigure 3.8, used in this thesis as the advanced instrument.

30


FFigure 33..88:: Scanning electron microscope (JSM-6490LV), used for advanced

measurements.

This microscope can be used for two purposes. One is for taking the scanning electron microscopy (SEM) images with 10-10,000X magnification and second is for qualitative and quantitative chemical analysis by using an electron gun with 0-30KV. Firstly, the purpose of the chemical analysis will be presented and, then, the usage of magnified images.

According to CSDS and MSDS, the chemistry of the cathode is NCM. But there are different types of NCM chemistry, based on their molar ratio. Various combinations of molar ratio of nickel, manganese and cobalt are available. Some of the well-known combinations are NCM111, NCM424, NCM523, NCM 622 and NCM226.

NCM111 means that the molar contents of Ni, Co, Mn are equal and NCM523 means the molar contents are 50%, 20% and 30% respectively. Although there were several attempts to contact different offices of the producer, no answer regarding the cell’s chemistry was found. Because of this situation, a chemical analysis was performed to determine the exact chemistry of the cell. FFigure 3.9 shows the result of this study.

31


FFigure 33..99:: Analysis of the cathode materials.

In FFigure 3.9, X-axis is using keV (kilo-electron-Volt) as reference unit and Y-axis has cps/eV (count-per-second/electron-Volt) as measure unit. The graphical representation shows the intensity of the signal. Also the term “K-series” is referring to the signal in the area of 6keV to 8keV. For the signals below 2keV, the series is called “L-series”. In the same figure, the result of the chemical analysis is presented. For a better understanding of this study, the following abbreviations need to be explained:

unn. C [wt.%]: Raw data of element’s concentration norm. C [wt.%]: Normalized concentration to give a sum of 100 weight% Atom. C [at.%]: Atom concentration to give a sum of 100 atom% Error [wt.%]: The statistical error of the result (probably higher in reality)

For obtaining an accurate result, the test was repeated several times for different cathode samples and the all are similar to Figure 3.9. The closest known chemistry to theseresults is NCM424. This type of NCM is not available in the material database of the software, but there is an option included, which offers the possibility of adding a new material. In this case, six parameters are required:

1. Molecular weight (g/mol) 2. Density (g/cm3) 3. First charge capacity (mAh/g)

32


4. First discharge capacity (mAh/g) 5. Maximum open circuit potential (V) 6. Particle size (µm)

Molecular weight will be calculated based on the chemical analysis in the next paragraph. Maximum open circuit potential of most of the cathode chemistries like NCM111, LCO, LMO and NCA are 4.3V, according to material database of the software. After consulting with the technical support team of the software, the value of 4.3V was taken into consideration for different NCM combination. The topic of particle size will be presented in section 3.2.2.2. The other three parameters, density and first charge/discharge capacities need to be estimated or measured. In one of the research papers published by the technical support team of the software, these parameters are calculate for NCM523. Due to the fact that there is not a big difference between NCM424 and NCM523, the values determined in this paper were used for these three parameters. Obviously, this can be considered as a source of error, because these parameters have impact on the designed model. Briefly a new material called “NCM523” added to the material database of the software. Also for checking how much this can affect the results, it is decided to have two models, one with chemistry NCM111 and another one with chemistry NCM523 and the names of the models are called based on their cathode chemistry “electrochemical NCM111” and “electrochemical NCM523”. Molecular weight or molecular mass is defined as the mass of one molecule. It is calculated as the sum of the mass of each constituent atom multiplied by the number of atoms of that element in the molecular formula. The atomic weight or mass of each atom is taken from the periodic table. The molecular formula of NCM111 is LiNi Co Mn O , where == = 0.3333%. For the model “electrochemical NCM111”, the calculation of molecular weight is presented in TTable 3.1. For the model “electrochemical NCM523”, the molecular weight is calculated in Table 3.2 based on the chemical analysis showed in Figure 3.9.

33


TTable 33..11:: Calculating the molecular weight of LiNi0.3333Co0.3333Mn0.3333O2. Atomic weight

(u) # of atoms in

molecular Total mass

(g/mol)

Nickel 58.6934 0.333333 19.5644471

Cobalt 58.933195 0.333333 19.6443787

Manganese 54.938045 0.333333 18.3126634

Oxygen 15.9994 2 31.9988

Molecular weight: 996.4612891

TTable 33..22:: Calculating the molecular weight of LiNi . Co . Mn . O .

Atomic weight (u)

# of atoms in molecular

Total mass (g/mol)

Nickel 58.6934 0.42 24.651228

Cobalt 58.933195 0.18 10.6079751

Manganese 54.938045 0.4 21.975218

Oxygen 15.9994 2 31.9988

Molecular weight: 996.1742211

The chemical composition and the molecular weight are determined using an electron microscope, as it is shown above. Thickness of the layers, particle’s size and porosity are three important topics for which the electron microscope and magnified images can be used. In this section, the measuring of the layers’ thickness will be presented, but the topics of the particle size and porosity will be covered in section 3.2.2.

Measuring the thickness of coated electrodes, foils and separator are crucial for a good model. As showed in section 3.2.1.2.2, simple instruments may do not have enough accuracy for measuring the thickness of thin layers like copper foil which need at least 1µm accuracy. In these cases, the microscope can be used for measuring the thickness of thin layers. Moreover for measuring the thickness of the foils in the coated area, the simple instruments do not have any chance and we forced to use the microscope.

For measuring the thickness of the coated materials/foils, the area near the edge of the plates is not good for taking probes, because the coated materials cover the foils. This can be seen by comparing FFigure 3.10 with Figure 3.11.

34


FFigure 33..110:: A SEM image from the edge of a cathode plate.

FFigure 3.10 shows the thickness of a cathode plate without any preparation or cutting. The aluminium foil is completely covered with the coated material and it is not possible to see or measure it. The cutting and preparing of the samples is very important step in the measurement process. The coated materials and also the thin foils have a soft structure, which makes the preparation process even more difficult. Cutting by blade and scissors are two simple ways for preparing the probe. The microscope images proved that cutting the samples by scissors had led to better results than using blades. The scissors are better for cutting the foil, because the force orientation is double-sided and comparing to blades, which has a one-sided force orientation, they produce less bending. Figure 3.11 presents the thickness and structure of an anode plate cut by scissors.

Figure 33.111: Thickness of an anode plate (cut by scissors).

35


The sample showed in FFigure 3.11 was cut with sharp scissors, but still the copper foil in the center of the coated materials does not have a uniform thickness and this makes the thickness measurement challenging even when an electron microscope is used.

The position and angle of the sample on its holder could also affect the results of the thickness measurements. Figure 3.12 presents how the cutting of the sample is influenced by these aspects.

(a) (b)

Figure 33.112: Thickness of two graphite plates, (a): Sample located below its holder, (b): Sample located above its holder.

The sample showed in Figure 3.12(a) is located a little bit lower than the sample holder and this is causing some part of the image to be covered by the sample holder, but this is not happening in the sample showed in Figure 3.12(b). Also, by putting the sample higher than the holder, it can be seen if it is perpendicular to the microscope or not.

Another important aspect is that the samples should not be pressed or scratched. Also, there is the situation, when some cracks appear between the coated materials and the foils, which can change the results of the measuring. For a good image of the samples, the microscope has to be set correctly (internal settings). After all the stages of the preparing process were fulfilled, the thickness of the layers was measured from different samples. The results of the measurements with simple and advanced instruments for the layers' thickness are presented in the Table 3.3 and they are compared with the paper [26].

36


TTable 33..33:: The results of the thickness measurements.

Parameters Unit Simple

instruments Advanced

instruments Paper [26]

Aluminium foil m 20 20 21

Cathode coated material (1 layer) m N/A 67-71 70

Cathode plate (3 layers) m 150-168 157-161.5 161

Separator m 20 25.5-27 25

Copper foil m 10 10-12.8 12

Anode coated material (1 layer) m N/A 76-81 79

Anode plate (3 layers) m 170-190 165-171.4 170

In TTable 3.3, the results of measuring with advanced instruments are shown in a range, because they are not the same for different plates and different samples, so instead of a value, a range is given. This situation is also similar for the simple instruments, but since their accuracy is not so good, then the given range is bigger.

In conclusion, the advanced instruments are more capable and more accurate in measuring the thickness of layers, but still some errors exist in the measurements (mostly because of the sample preparations), which should be considered in the future works. For the designed electrochemical model, 10 parameters (7%) of the 138 parameters are measured with advanced instruments. Next section presents the third source of data.

3.2.1.3 Physical tests (tuning the parameters)

This source is used in the black box method, intensively, for creating the ECM, but as mentioned before, gray box method uses this source, partially. Physical tests cannot be used to extract the parameters' values of electrochemical model directly, except of three parameters, contact resistance of two electrodes and OCV@100%SOC (these parameters are covered in section 3.2.2). Then the way for using the physical tests is different and that is tuning.

After designing the cell and considering the measured and estimated parameters from the two previous sources, a solution for reducing the errors is tuning the parameters' values based on the physical tests. The results of some physical tests like 1C-discharge and HPPC on a physical cell can be considered as reference and by checking the difference between the output of the designed model and the reference (the error), a closed loop controller can be used to change the model's parameters. This can highly compensate the uncertainty of predicted parameters and reduce the errors.

37


Tuning the parameters is suggested by the AutoLion, too. This is presented in a video clip, published in their YouTube channel. Two screenshots from this video, before and after the tuning are shown in FFigure 3.13.

(a) (b)

Figure 33.113: Two screenshots from the AutoLion video clip, (a): before tuning and (b): after tuning the parameters’ values.

Figure 3.13(a) shows the results of the model designed by the software and physical test for the 1C-discharge and HPPC tests before the tuning process. Figure 3.13(b) shows the same tests, but with improved results for the designed model after the tuning process. The tuned parameters, showed in the video clip, are the below 7 parameters:

2-Thicknesses of electrodes plates 1-Loading of cathode 1-Particle size of anode 3-Bruggeman exponents

The tuning process offered by the software requests changing the parameters' values manually, which is time consuming and, sometimes, may not reach to the best results. In this case, designing a tool like a closed loop controller for tuning the parameters' values is needed and it can be based on the two mentioned physical tests.

38


The first physical test, named 1C-discharge test, is always performed first in the series of physical tests. This test can check the cell’s health and capacity before running the other expensive and time consuming tests. This test takes one working day and, also, the results of this test are available in most of CSDSs. The second test is the standard HPPC test, which can determine the dynamic power capabilities of the designed model.

A simple way for tuning the parameters is fixing all of the parameters except one. Then the value of this one parameter changes in each simulation test. This procedure is repeated for each value and each parameter, but this is a time consuming procedure. For example (with underestimating), if just 10 parameters are needed for tuning and just 10 values for each parameter can vary, the number of simulation tests are 10 , which equals to 10,000,000,000. If two mentioned simulation tests takes 10 seconds (underestimated), the tuning process will take more than 3 years with a normal computer. It is obvious that an intelligent method is needed to reduce the number of simulation tests. This task is challenging due to the high number of parameters and variables.

Since this task is a challenging task and it is out of the time frame of this thesis, it was just mentioned. Without any doubt, this is an important part of the modeling process and it should be part of the future works. In the overall, two electrochemical models NCM111 and NCM523 are designed without tuning their parameters.

It is worth to mention, the intended tool, besides reducing the error of the model, it is also useful in identifying the influence of the parameters. The ones with the highest impact are called “key parameters” in this thesis. In the next section, key parameters are explained comprehensively.

3.2.2 Key parameters

Key parameters are influencing the results of the model considerable. The values of some of the key parameters can be measured with simple/advanced measurements like the dimensions of the plates or the thicknesses of them. They are called “key-measured” parameters. For this type of the key parameters, a reliable value can be achieved with no or low cost in resources. But measuring the values of some of the parameters are difficult and costly and therefore their values need to be estimated, like the loading of the electrodes. The values of this type of the parameters which are called “key-estimated”, should be selected precisely. Studying these parameters comprehensively gives us the ability to estimate them precisely.

In this section, the key-estimated parameters are presented. Until now, 12 parameters were identified as key-estimated parameters:

2-Loading of electrodes

39


2-Particle size of electrodes 1-OCV@100%SOC 2-Contact resistance of electrodes 2-Solid state diffusion coefficient of electrodes 3-Bruggeman exponents

In the rest of this section, these 12 parameters are discussed in detail.

3.2.2.1 Loading

Active material loading [mAh/cm2], also called “aerial capacity”, for both the cathode and the anode constitutes two parameters in the electrochemical model. These parameters are a representative of the energy density [Wh/L] and the specific energy [Wh/Kg]. Also their value can be increased arbitrarily by increasing electrodes thickness and/or density.

Moreover, based on our experiments during the designing of the model, these two parameters have a key role in the capacity and weight of the cell. In TTable 3.4, the values of the cathode loading are changed to see the effect of it on the cathode porosity, capacity and weight of the model.

Table 33.44: The effect of the cathode’s loading values on capacity and weight of the designed model.

Cathode Loading (mAh/cm2)

Anode Loading (mAh/cm2)=

Cathode Loading * 115%

Reversible capacity (mAh)

Cathode Porosity (N/A)

Total Weight (g)

3.9000 4.4850 30,000.4343 0.1700 460.5500

3.8000 4.3700 29,231.1924 0.1900 455.3000

3.7000 4.2550 28,461.9505 0.2100 450.0500

3.6000 4.1400 27,692.7086 0.2300 444.8100

3.5000 4.0250 26,923.4667 0.2600 439.5600

3.4000 3.9100 26,154.2248 0.2800 434.3100

3.3000 3.7950 25,384.9829 0.3000 429.0600

3.2000 3.6800 24,615.7410 0.3200 423.8100

3.1000 3.5650 23,846.4991 0.3400 418.5600

3.0000 3.4500 23,077.2571 0.3600 413.3100

2.9000 3.3350 22,308.0152 0.3800 408.0600

40


In TTable 3.4, the value of anode loading changed automatically and it is equal to 115% of cathode loading’s value. The ratio of 115% can be defined by the user in the parameter “N/P ratio”. From Table 3.4, it can be concluded that, by decreasing both cathode and anode loadings, the capacity and weight of the model are decreasing too, but porosity of electrodes is increasing. The porosity of electrodes are not a part of the electrochemical model’s parameters, but, by measuring the porosity, the loading values can be selected precisely. In the rest of this section, the porosity of the electrodes are discussed.

3.2.2.1.1 Porosity

The definition of porosity is, a measure of the void/empty spaces in a material, and is a fraction of the volume of voids over the total volume, between 0 and 1, or as a percentage between 0 and 100%. It is defined by the ratio:

= (3.1)

where is the volume of void-space and is the total or bulk volume of material, including the solid and void components.

There are several methods for measuring the porosity of materials. They can be divided into two categories: imbibition and optical methods. In the first type, a liquid or gas is used to fill the voids and then the volume of the liquid/gas is measured as the volume of porosity. For example oil absorption number (OAN) is a method for calculating the porosity of black carbon by using the oil (for more information the readers are referred to the ISO 4656:2007). Water saturation, mercury intrusion porosimetry and gas expansion are also some other methods in the imbibition type. In the second type, the voids are not filled, but instead they are identified by the microscope or computerized tomography (CT) scanner. Using an industrial CT scanning for creating a 3D rendering of the external and internal geometry, including the voids can be a reliable method for measuring the porosity.

In the case we did not have access to the tools for the mentioned methods, then a method with less reliability can be employed instead. Taking 2D images by microscope and analyzing the images could give an estimation for the porosity. This estimation is expected to be imprecise, because 3D properties cannot be measured correctly by 2D images. In this way, the magnification rate, the contrast and brightness of the image and also the location of the selected samples for study have a big impact over the result. These factors should be considered before measuring the porosity with 2D images. As an example, the effect of the samples location is presented in the next paragraphs.

Figure 3.14 shows two images from the anode materials taken by the microscope. Figure 3.14(a) shows a part of the anode, which cuts from the surface of the anode in contact with the copper foil. The big and small voids are easily detectable in this image.

41


In contrast, FFigure 3.14(b) shows the surface of the anode in contact with the electrolyte and separator. In this image except of very small holes and cracks, there is not any void. This can be seen by comparing two orange circles.

(a) (b)

Figure 33.114: Two different samples from the anode. (a): An anode sample

scratched from the plate. (b): An anode sample from the surface of the plate.

The best explanation for the difference between these two samples is that voids are filled by lithium salt. The images were taken at least two weeks after opening the cell. The electrolyte, starts to evaporate immediately after opening the cell, but some amount of the ingredients, like the lithium salt are remaining on the surface of the electrodes. They fill the voids and during the natural drying process some small cracks are created.

Both samples showed in Figure 3.14 are from the same plate, which is located in the middle of the other plates. The position of the plate is very important, since the first and last ones have more voids filled with lithium salt than the ones in the middle. Figure 3.15 shows the difference between the two different anode plates. Figure 3.15(a) is the first plate in the top and Figure 3.15(b) is a plate from the middle. Also their colors are different, the plate showed in Figure 3.15(a) is brighter comparing to the plate showed in Figure 3.15(b). Moreover in Figure 3.15(a) some bubbles with different dimensions can be seen, which do not exist in the Figure 3.15(b).

42


(a) (b)

FFigure 33..115:: Two anode plates. (a): The first plate in the top. (b): The middle plate.

For understanding what is in the surface of the anode plate in FFigure 3.15(a) a material analysis was run and the results are shown in Figure 3.16. Four elements are detected, carbon (C), oxygen (O), fluorine (F) and phosphorus (P). Fluorine and phosphorus are the elements of the lithium salt. The signal of fluorine is around 5.3 times more powerful than the signal of phosphorus. This is coincide with the lithium salt formula, LiPF6 which contains 6 atoms of fluorine and 1 atom of phosphorus. Also, the signals of lithium salt elements are more powerful than the signal of carbon. It means, most of the surface was covered with the lithium salt elements. In contrast, the material of the anode plate, showed in Figure 3.14(a), was studied and the results are shown in Figure 3.17.

43


FFigure 33..116:: Material analysis of the first anode plate.

FFigure 33..117:: Material analysis of the anode plate showed in FFigure 33.14(a).

The detected elements in FFigure 3.17 are the same as Figure 3.16, plus one more element, copper (Cu). The results are confirming the statements from the previous sections, because the copper is detected due to the fact that the studied sample was cut from a side of the plate connected to the copper foil. Fluorine and phosphorus are detected again, but much less than the Figure 3.16. Still the signal of fluorine is more powerful than the phosphorus, which means the elements are coming from the lithium salt. The decrease of

44


lithium salt elements in the two studied samples is significant and can confirm the mentioned presumption that the voids in FFigure 3.14(b) are filled with the lithium salt materials.

In addition, the signal of oxygen also was decreased significantly and this may explain the amount of reactions between the graphite and electrolyte materials in the two opposite sides. Also it is worth to mention the Li element is not detectable with our electron microscope.

Briefly, two different samples of anode materials were studied. One, the surface connected to the electrolyte/separator of the first plate of the anode showed in Figure 3.15(a) as the most area with the filled voids and another one, the surface connected to the copper foil of a middle plate of the anode as the least area with the filled voids. The material analysis shows a big difference for the lithium salt materials between these two samples. The selecting of the samples has a big effect over the porosity value. Also the result of this study should be considered for measuring the electrodes porosity with the two mentioned methods, imbibition and optical. The samples for studying the porosity in this thesis are selected from the center of the middle plates on the side connected to the foils. In the next three sub sections, the porosity of cathode, anode and separator is presented.

3.2.2.1.2 Porosity of the cathode

In this section, the porosity of the cathode is presented and based on the calculations, a range for the porosity value is suggested. In the above sections, it was proven that some effective factors should be considered first. The samples have to be selected from the side of plate connected to the foils. The magnification rate for the images is around X400-X1500, which can ensure the visibility of enough details. Too less magnification like X100 cannot detect the voids and too much magnification like X10000 just shows a single particle/void. The brightness and contrast of the images are modified in the microscope's settings for sharp and lucid images.

NCM has spherical particles which can be identified and detected easily. In this thesis, a simple method was employed for calculating the porosity. Firstly, the images from the microscope were converted to grayscale images and then a threshold technique was used for converting them into binary images. Finally, the number of black pixels were counted as voids.

Selecting a right level of threshold is very important for counting the voids. The threshold level can take different values for some images, so a judgment with the naked eye was used to select the right threshold level. To compensate the human error, a range of the threshold levels was selected. Table 3.5 shows the calculated porosity for three different

45


magnification levels, X1500, X1000 and X400 and also the selected suitable range of the threshold levels.

TTable 33..55:: The measured porosity of the cathode for different magnification rates and different threshold levels.

Threshold level 30 35 40 45 50 55

X1500 21.29% 23.51% 29.40%

X1000 25.22% 25.31% 27.14%

X400 25.27% 30.57%

From TTable 3.5 we can conclude that the porosity of the cathode is in the range of 21%-30%. Figure 3.18 shows two cathode images with X1000 magnification and 45 threshold level. Figure 3.18(a) is the image taken by the microscope and Figure 3.18(b) is the analyzed image after applying the filter and threshold technique. Black pixels in Figure 3.18(b) were counted as the voids. The total number of the black pixels divided by the total number of the whole black and white pixels gives the porosity value, as it showed in equation (3.1). Multiplying with 100, gives the percentage of porosity.

(a) (b)

Figure 33.118: The cathode material with X1000 magnification. (a): The image from the microscope. (b): The same image after applying the threshold technique, at level

45.

46


FFigure 3.18 shows the porosity around 25% which seems to be the most reasonable value for the cathode porosity.

3.2.2.1.3 Porosity of the anode

Calculating the porosity for the anode is more difficult than for the cathode and, consequently, less accurate. It is due to the flaky structure of the particles. The same three above mentioned factors: sample locations, magnification rate and contrast/brightness were considered for measuring the porosity of the anode.

For the level of magnification, just two rates X1000 and X400 were studied because the size of the anode particles are normally around 150% bigger than the cathode-NCM particles, which means that the X1500 magnification rate cannot show enough particles/voids and it was disregarded for this study. Except of the magnification rate, the same procedure as in the cathode case was repeated for the anode and the results are shown in the Table 3.6.

Table 33.66: The measured porosity of the anode for different magnification rates and different threshold levels.

Threshold level 105 110 115 118 120 125 130 135

X1000 28.06% 29.50% 29.84% 32.45% 36.21% 40.81%

X400 36.46% 40.67% 40.82% 43.27%

By comparing the Table 3.5 and Table 3.6, it can be concluded, that the porosity of the anode (graphite) is higher than the cathode (NCM). Then the anode porosity should be in the range of 28%-43%. Also the results showed in Table 3.6 may be over-estimated, because as mentioned, the graphite has a flaky structure which makes the detection of voids difficult especially in 2D images. Figure 3.19 presents an anode sample with X1000 magnification.

47


(a) (b)

FFigure 33..119:: Anode material with X1000 magnification. (a): the image from the microscope. (b): the same image after applying the threshold technique at level 118.

FFigure 3.19 shows a porosity of 32%, which is the most reasonable value for the porosity of anode.

3.2.2.1.4 Porosity of the separator

Separator is a critical component in LIBs. The structure and properties of the separator are considerably affecting the battery performance, including the battery's energy and power densities, cycle life, and safety. Also, the separator must have sufficient pore density to hold liquid electrolyte that enables ions to move between the electrodes. Excessive porosity hinders the ability of the pores to close, which is vital to allow the separator to shut down in the case of an overheated battery.

A separator generally consists of a polymeric membrane forming a microporous layer. The voids of the separator are in Micro or Nano meter size. Then, porosity of the separator is not calculated in this thesis due to the incapability of the microscope for taking the image of the separator's voids. Figure 3.20 shows two images of the separator with two magnification rates, X2000 and X500 and in two positions, vertical and horizontal. Even with 2000 times magnification, detecting the voids is impossible, as it can be seen in Figure 3.20(a).

48


(a) (b)

FFigure 33..220:: The separator layer under the microscope. (a): In vertical position with X2,000 magnification. (b): In horizontal position with X500 magnification.

As mentioned in section 3.2.1.1.3, the typical value for separator’s porosity is 40% and this value was used in the electrochemical model. In conclusion, for measuring the porosity of the separator, cathode and anode, employing the mentioned reliable methods is suggested as one of the most important task for future works.

3.2.2.2 Particle size

Particle size [µm] or particle radius of cathode and anode are two parameters in the electrochemical model, which influence both capacity and Coulomb efficiency. Reducing the particle size will increase the specific surface area and change some of the important battery characteristics like capacity and the void spaces between the electrode particles. In [27], it is found that the conductivity and diffusion coefficient is increased by an order of magnitude, and the activation energy is reduced by half if the particle size reduced from micro-level to nano-level.

Size of the particles cannot be completely homogeneous, so a graph represents the distribution of particle size. In FFigure 3.21, the particle size variations for the cathode of the sample cell can be seen clearly. The green arrows show the small particles and the orange ones, the big particles.

49


FFigure 33..221:: A SEM image of the cathode plate.

Due to the variation of particles size, showed in FFigure 3.21, instead of a value, the size of the particles is shown alongside their distribution. Figure 3.22 is an example for presenting the particle size distribution (PSD).

Figure 33.222: The particle size distribution curve of a NCM chemistry, from

paper [28] (by varying different content of Ti-doping).

50


FFigure 3.22 shows the PSD of NCM chemistry for a cell in paper [28]. The default value for NCM particle size in the software is 10µm and Figure 3.22 as an example, confirm the same value. In this thesis, the required instruments to do such analysis were not available, but maybe a SEM image taken from the sample cell shown in Figure 3.23, could prove that the value of 10µm is a good estimation for the cathode particle size.

Figure 33.223: Particle size of the NCM chemistry taken by the microscope from the

sample cell.

The SEM image in Figure 3.23 was taken from the surface of the cathode and the orange lines have the same size as 10 m scale line. The size of some of the particles cannot be seen completely in the 2D image and then by considering this point, it can be said, 10 m is the mean value of our sample cell too.

The situation for measuring the particle size of graphite is worse. As mentioned in section 3.2.2.1.3 and showed in Figure 3.19, graphite particles have a flaky structure and this makes the particles indistinguishable from each other. The default value for graphite particle size is 15µm. The SEM images taken from the graphite are not useful for estimating this value and, then, the default value was used in the model.

The particle size can be measured by X-ray powder diffraction, particle size analyzer and scanning electron microscopy. Nano particle analyzers from HORIBA company are well known in this field of work.

51


3.2.2.3 Open circuit voltage at fully SOC

The open circuit voltage [V] at fully SOC (OCV@100%SOC) is one of the parameters in the electrochemical model, which can be found in the initial conditions tab. In this thesis, this parameter is considered as a key parameter due to the effect of its value over the results. This is acquired based on the performed experiments during the modeling process.

This parameter defines the OCV at fully SOC and consequently the voltage of the OCV curve. The OCV should be correct in order to reach a good accuracy in the other tests like HPPC test. The values of these parameters are important for obtaining an accurate model in the overall. One of the tests in the verification section is OCV. More details are given about this parameter and its effect over the model's results in section 4.3. For measuring this parameter, a simple physical test is required.

In this thesis, different values were selected for this parameter to check the influence of it over the models. As an example for the electrochemical model NCM111, two values 4.2V and 4.0V were selected and the results are shown in FFigure 3.24, Figure 4.10 and Figure 4.11.

3.2.2.4 Contact resistance of electrodes

Electrical contact resistance (ECR) [ m ] at the interface of both electrodes and current-collector bars in LIB assemblies are two parameters in electrochemical model. ECR is a direct result of contact surface imperfections, i.e., roughness and out-of-flatness, and acts as an ohmic resistance at the electrode–collector joints. The effective conductivity of lithium-ion battery electrodes has a strong impact on their performance and usable capacity. In [29] it is showed that the energy loss due to ECR in the considered LIB, can be as high as 20% of the total energy flow in and out of the battery under normal operating conditions. However, ECR loss can be reduced to 6% when proper joint pressure and/or surface treatment are used. A poor connection at the electrode–collector interface can lead to a significant battery energy loss as heat is generated at the interface. Consequently, a heat flow can be initiated from the electrodes towards the internal battery structure, which results in a considerable temperature increase and onset of thermal runaway. At sever conditions, heat generation due to ECR might cause serious safety issues, sparks, and even melting of the electrodes [29].

Like the parameter OCV@100%SOC, contact resistance is also considered a key parameter during the modeling and verification process of the models. Figure 3.24 shows the effect of contact resistance’s value in resulted model. In this figure, the drop of the voltage is caused by the contact resistance.

52


FFigure 33..224:: The influence of the two parameters: OCV@100%SOC and contact

resistance on the results of the voltage test.

In FFigure 3.24, the dropped voltage for physical test on the cell id 32 is 0.203V, but for the ECM is 0.074V. The higher dropped voltage for the physical test is due to the fact that there was any welding of the cables with the electrode–collector joints. The dropped voltage for the ECM is more realistic than the physical test. Therefore 0.074V is selected as reference for designing the models.

The default value 0.0002 m2 for these parameters leads to 0.074 drop voltage, which is the same as the selected reference. More explanation about this figure are given in chapter 4.

In [29], a study about this parameter are done by analyzing the three major components: surface topology, contact mechanics and electrical transport. Similarly to OCV@100%SOC, for measuring the contact resistance, a physical test is required. Also in [29], a novel method for measuring the contact resistance of electrodes is given.

3.2.2.5 Solid state diffusion coefficient of electrodes

Solid state diffusion coefficients of anode and cathode [m /s] are also known as solid phase Li diffusion coefficients and they represent two parameters in the electrochemical model. In some papers, they are denoted as D . Also D is used for electrolyte phase Li+ diffusion coefficient. In order to determine the important characteristics for the charge regimes (current and time), the rates of electrochemical Li insertion have to be known.

53


As far as the rate of intercalation/deintercalation is limited by Li+ ion diffusion, the Li+ diffusion coefficient value, along with the thickness, is eventually responsible for these parameters [30].

Some methods exist for estimating the values of the diffusion coefficient. The most common way is measuring D from the easily measurable electric conductivity, which according to [30] is not reliable. In the same paper, a more reliable method presented, which is based on potential decay curves after the current is being switched off. Also these parameters are dependent on temperature, SOC and current density. Moreover, while the cell ages, the diffusion coefficients decrease as well [31].

In this thesis, these key parameters are not studied completely and the default values were used in the model. The default values of these parameters are not visible for the users in the software, but in [32] for their cell, the values are 2.0 10 2/s for the anode and 3.7 10 2/s for the cathode.

3.2.2.6 Bruggeman exponents

Bruggeman tortuosity exponents [no unit] for electrolyte diffusion in cathode, anode and separator are three parameters in the electrochemical model. According to [33], for more than 50 years, porous electrode theory has been relied on the use of equation (3.2), as proposed originally by Bruggeman to estimate the electrode tortuosity.

= 1 (3.2)

where is the tortuosity and is the porosity. Bruggeman's exponents yield an important quantitative measure, the connectivity, to characterize the physical path for transport through the underlying phases [34].

In this thesis, these key parameters were not studied completely and the value 1.5 which is typical, was considered for the model. But in paper [35], it is showed that a Bruggeman exponent of 1.5 is often not valid for real electrodes or separator materials. Also it is found that only idealized morphologies, based on spherical or slightly prolate (i.e. rod-type) ellipsoids, are expected to give rise to a Bruggeman law with an exponent of about 1.3. Porous networks based on other particle morphologies such as oblate (i.e. disk-type) ellipsoids or lamellar or flaky materials increase the tortuous path for ionic conductivity and result in a significant increase of the Bruggeman exponents.

A way for calculating these exponents is performing numerical diffusion simulations on three-dimensional microstructures, obtained from tomographic techniques. But in paper [36] from year 2014, a tool called “Bruggeman estimator” is presented, which can determines the Bruggeman exponents based on the differential effective medium

54


approximation that just requires two microscope images: one of the top and one of a cross section through an electrode. As this tool just requires two microscope images as input, then it should be considered in the future works for estimating the value of Bruggeman exponents.

3.2.3 Analyzing the parameters

Two electrochemical models NCM111 and NCM523, were created by the use of the gray box method. The value of the model’s parameters were taken from two sources: knowledge-based and physical/chemical measurements. The third source, physical tests were not used in this thesis for the parameters' tuning.

The values of some parameters are available in more than one source. For example the dimensions of the cell are available in CSDS, papers (knowledge-based source) and in simple measurements (physical/chemical measurements source). From the different sources, the most accurate values were selected for the parameters' values.

Two electrochemical models NCM111 and NCM523, are similar to each other, except in the cathode chemistry case. Due to their similarity, instead of the both models, the parameters of the NCM523 will be analyzed in this section.

3.2.3.1 Sources of the parameters

The usage of each source for creating the model is shown in FFigure 3.25. The Knowledge-based source has the most contribution with 55% of all the 138 parameters. This portion for the knowledge-based source, which is mostly free, is part of the main idea behind the gray box method. The physical/chemical measurements source is contributing in 20% of the parameters and the rest of the parameters (25%) were not applicable for the model. The physical tests (tuning the parameters) as mentioned before were not used in this thesis.

55


FFigure 33..225:: The usage of each source for creating the NCM523 electrochemical

model.

Each of the sources is divided into some groups. The details of each sources and their contribution to the model are shown in FFigure 3.26.

Figure 33.226: The usage of each source for creating the NCM523 electrochemical

model (in detail).

The material database and its default values from the software have the most usage to the model with 46% in the knowledge-based source. In the physical/chemical

55%

20%

0%

25%

SOURCES OF PARAMETERSKnowledge-based source Physical/chemical measurements source

Physical tests (tuning the parameters) Not applicable/control parameters

46%

2%7%7%

13%

11%

14%

SOURCES OF PARAMETERS (IN DETAIL)

Material database of the software CSDS & MSDS

Papers, journals, books Advanced instruments

Simple measurements Control parameters

Not applicable for this model

56


measurements source, the simple measurement with 18 parameters (around 13%) and the advanced measurements with 10 parameters (around 7%) contributed to the models.

In the overall, if the advanced measurements are considered as expensive measurements in the modeling process, then just 7% of the parameters need to use these costly and time-consuming tests. The rest of them can be acquired with low or no cost from this point of view. In conclusion, the gray box method could overcome the obstacles of cost and time for developing the model.

3.2.3.2 Parameters’ influence over the model

The model’s parameters can be divided into two groups of key and not-key parameters. Also as mentioned in section 3.2.2, the key parameters can be divided into two groups: measured and estimated. The portion of each of these groups are shown in FFigure 3.27.

Figure 33.227: Parameters’ influence over the NCM523 electrochemical model.

Around a half (46%) of the parameters are “not-key” parameters. The values of these parameters were estimated with an acceptable accuracy as they are not influencing the results too much. The key parameters are 29% of the parameters, which contain measured and estimated parameters. Just 9% of the whole parameters are the key-estimated parameters, which need to be estimated carefully and with high precision. These parameters are explained in detail in section 3.2.2.

20%

9%

25%

46%

Key-Measured parameters Key-Estimated parameters

Not applicable/control parameters Not key parameters

57


3.3 MATLAB generic battery model

CSDS contains some useful and free information about the cell. To check if these information are enough to create an accurate model, different models were studied and, finally, a MATLAB model, called “generic battery model”, was found. This model is available in Simscape of MATLAB 2013 or higher versions. Simscape is similar to Simulink, but, in Simscape, the physical elements with wire connections are used. All elements have to be connected to each other through the wires by considering the + or - poles.

The input parameters for this model are 9 parameters which are listed and explained in TTable 3.7. They are mainly based on the discharge characteristic curve, available in CSDS. The description of this model is available in the website of Mathworks.

Table 33.77: The list of the MATLAB model’s parameters.

# Parameter Unit Description of parameter

1 Nominal voltage V The nominal voltage represents the end of the linear zone of the discharge characteristics

2 Rated capacity Ah The rated capacity is the minimum effective capacity of the battery not the nominal capacity

3 Maximum capacity

Ah The maximum theoretical capacity. This value is generally equal to 105% of the rated capacity

4 Fully charged

voltage V The fully charged voltage for a given discharge current. Not the no-load voltage!

5 Nominal

discharge current A

The nominal discharge current, for which the discharge curve has been measured

6 Internal

resistance Ω

Internal impedance at 1kHz available in all of the CSDSs

7 Capacity @

nominal voltage Ah

The capacity extracted from the battery until the voltage drops under the nominal voltage

8 Exponential zone, voltage

V The voltage corresponding to the end of the exponential zone

9 Exponential

zone, capacity Ah

The capacity corresponding to the end of the exponential zone

The Simulink/Simscape file contains a generic battery model and the input/output blocks are depicted in Figure 3.28.

58


FFigure 33..228:: The Simulink/Simscape file contains the MATLAB model.

FFigure 3.28 represents the input, the output and the generic battery model. The values of the model’s parameters are listed in Table 3.8. These values were acquired based on CSDS of the sample cell.

Table 33.88: Parameters’ values of the MATLAB model.

# Parameter Unit Value

1 Nominal voltage V 3.4

2 Rated capacity Ah 20

3 Maximum capacity Ah 21

4 Fully charged voltage V 4.15

5 Nominal discharge current A 6

6 Internal resistance Ω 0.0012

7 Capacity @ nominal voltage Ah 18

8 Exponential zone, voltage V 4.1

9 Exponential zone, capacity Ah 0.5

This model can predict the voltage and SOC of the model. Also the temperature can be taken into consideration if the discharge curve is available for different temperatures.

As a summary, two electrochemical models NCM111 and NCM523 were created by using the gray box method. Two out of the three sources, knowledge-based data and physical/chemical measurements were used for extracting the values of the models. The third source, physical tests which can be used for tuning the parameter’s value was not used, due to the limited time in the thesis. Besides the electrochemical models, a

59


MATLAB model, called “generic battery model”, was created based only on the information from the CSDS.

60

Chapter4. Verification

In this chapter, the results of the designed models are compared with different references. These tests will help us to understand how much the accuracy of the proposed method is close to the desired method.

To examine the accuracy of the designed models, some references are needed. The available references for this thesis are, two physical tests run on two cells with id numbers 32 and 33, ECM based on physical tests on a cell with id number 35, CSDS of the sample cell and also some open source data like the research documents or papers which presents physical tests on the same type as the sample cell.

Although the three physical cell ids 32, 33 and 35 are from the same manufacturer and have the same type, but probably they are produced in different manufacture lines, batches and dates.

The date of the tests on cell ids 32, 33 and 35 are March 2009, February 2010 and, respectively, October 2010. This has the meaning that the tests are done in different dates and, maybe, the cells are from different versions. This is unclear and, then, it should be considered when the results of these cells are compared with the results of the other references/models.

In this thesis, four tests are used to analyze the behavior of the models in different aspects. The tests are:

Capacity test

44 VVerification

61


Voltage during the 1C continuous discharge test OCV test HPPC test

These four tests are the most well-known tests for lithium-based battery and their results present some of the characteristics and behavior of the battery. Not all of the four tests have all the available references, but, during the verification, the references are used as much as possible.

4.1 Capacity test

The first test presented in the scientific literature concerning testing of the lithium based batteries is the capacity test. The capacity can be used for the other tests too or in the comparison of models based on their capacity. The calculated capacity in this test will be used for the next three tests. In literature, the term “capacity” has different definitions. Some of these definitions are, also, used in this thesis and their explanations are given in below.

NNominal Capacity (for a specific C-rate) is the coulometric capacity, the total Amp-hours available when the battery is discharged at a certain discharge current (specified as a C-rate) from 100 percent SOC to the cut-off voltage. Capacity is calculated by multiplying the discharge current (in Amps) by the discharge time (in hours) and decreases with increasing C-rate.

Theoretical Capacity is the capacity based on the entire Li content within an active material, which not cycled between stoichiometry of 0-1, and therefore, the theoretical capacity, , may not be an accurate indicator of the material useable capacity. The theoretical capacity of an active material depends on its molecular weight and number of electrons that participate in lithiation or delithiation of the material.

Reversible capacity is the capacity of an active material based on the stable, cyclable Li after formation. In the other word, the reversible capacity is the capacity of the cell at infinitely slow discharge rate and normally is 10% higher than the nominal capacity.

Operational capacity is also the capacity of the cell at a very low C-rate and in the chosen voltage limits. This value can be the same or lower than the reversible capacity.

In this thesis the capacity test defined by the cell manufacturer, is used showing for charging the cell a CC/CV procedure with following parameters:

Charge: CC(0.5C)/CV(4.2V to 0.1C) Discharge: CC (1C) to 3.0V at 25

62


The procedure for standard capacity test is:

1. Constant current charge at 0.5C-rate until voltage reaches the cutoff voltage (4.2V in our case)

2. Constant voltage charge at 4.2V until current reaches the cutoff current (about the charging current)

3. Rest time (for the physical tests run on cell ids 32 and 33 are 30 minutes) 4. Constant current discharge at 1C-rate until the voltage reaches the discharge cutoff

voltage (3V in our case) 5. Rest time (for the physical tests run on cell ids 32 and 33 are 15 minutes);

Repeating this procedure for 3 times is called formation test. The Ah capacity of the third cycle is taken as the Ah capacity of the cell. In some literatures like [37] the rest time is 1 minute, but for the physical tests on cell ids 32 and 33 is 30 minutes. The procedure of the formation test on the cell id 32 is shown in FFigure 4.1(a).

(a)

(b)

Figure 44.11: The capacity test for the physical cell id 32. (a): The applied input

current and the voltage response. (b): The temperature of the cell during the test.

63


The temperature in all of the models is always fixed at the room temperature, 23 , but in the physical test it cannot be kept exactly 23 . According to the result of the physical test showed in FFigure 4.1(b), the temperature was in a narrow range between 21 and 24 which can be considered as room temperature.

4.1.1 References

For the capacity test, five references are available. This test has the highest number of references between the four tests presented in this thesis. The references are, CSDS, physical test on cell ids 32 and 33, ECM based on the physical test on cell id 35 and a research report based on the physical test on the same cell type.

4.1.1.1 CSDS

During the years, different CSDSs are published for the sample cell by the cell’s manufacturer. The physical tests on cell id 32 which is the sample cell in this thesis, were done in March, 2009. By considering the time length for running the tests, the cell shuld be produced in 2008 and hence a CSDS from that year was selected. According to the CSDS of the sample cell in the year 2008 the nominal capacity is 20Ah. The rate discharge characteristic of the CSDS is shown in Figure 4.2. This figure shows clearly, the measured capacity of the cell decreased by increasing the C-rate.

Figure 44.22: Voltage vs. capacity of the cell for different C-rates, taken from the

CSDS of the sample cell.

64


In FFigure 4.2 which is published by the manufacture, the label of X-axis is Ah. As a conclusion the 1C discharge capacity of the cell is considered to be 20.2Ah for the reference CSDS.

4.1.1.2 Physical tests on the cell ids 32 and 33

The formation test is available for the cell ids 32 and 33 at room temperature. One of the reasons for doing this test is extracting the capacity. The test is showed in Figure 4.1 for the cell id 32 and it is very similar for the cell id 33. For calculating the capacity, the amplitude of the applied current (A) is multiplied to the duration of the applied current in hour (h) for each step and then the sum of the results is considered as the capacity (Ah).

As it is shown in Figure 4.1, there are three cycles for each physical cells 32 and 33, and then all the results are shown in Table 4.1.

Table 44.11: The capacity test results for the two physical cells 32 and 33.

First cycle (Ah) Second cycle (Ah) Third cycle (Ah)

Physical cell 32 19.81 19.70 119.71

Physical cell 33 20.48 20.38 220.39

The third cycle is considered as the capacity of the cell for the 1C-discharge. The difference capacity between the two cell ids, 32 and 33 is 0.68Ah.

4.1.1.3 Equivalent circuit model

This model is an equivalent circuit model (ECM) with two RC-branches based on the physical tests on the cell id 35. Due to some technical problems, running the standard capacity test with the mentioned procedure was not possible for the ECM. Instead, the initial SOC of the model was set up to 100%SOC and 20A constant discharge current was applied to the model. By counting the time until the voltage of the cell reached to the low cutoff voltage (3V), and a simple multiplication of current (A) and time (hour), the capacity is calculable. Figure 4.3 shows the voltage response of the test from 100 seconds before applying the current until the end time of the test.

65


FFigure 44..33:: The ECM discharge curve.

For the ECM, the voltage reaches to the low cutoff voltage in 3627 seconds or 1.0075 hours, as it shown in FFigure 4.3. Then the capacity of the cell is 20A*1.0075h=20.15Ah. It is worth to mention, the SOC output port of the model shows 1.94%SOC when the cell reaches to 3V. The behavior of the model below the 2.8V is not reliable.

4.1.1.4 Research report

The research report [37] from California university contains some physical tests. The tested cell has the same type as the used sample cell, but it is possible that they have different versions and this has some impact over the results.

The mentioned procedure of “CC\CV” capacity test, explained in section 4.1, used in this research report for measuring the capacity. In their report, they did not mention the temperature for the test but it seems to be done at room temperature. The calculated capacity for the tested cell is 18.3Ah. It is worth to mention, in their tests, the rest times in CC\CV procedure, were just one minute.

4.1.2 Models

This sections shows the results of the three designed models, one MATLAB generic battery model and two electrochemical models. The explanation of the MATLAB model is given in section 3.3. Also the difference between the two electrochemical models are explained in section 3.2.1.2.3.

66


4.1.2.1 MATLAB generic battery model

The same procedure introduced in section 4.1.1.3, was repeated for the MATLAB model and this model showed a capacity of 17.89Ah for the cell. Also the SOC of the model showed a value of 14.79%SOC, when the voltage reached the low cutoff voltage (3V).

4.1.2.2 Electrochemical models

The same procedure introduced in section 4.1.1.3, was repeated for the electrochemical models, too. The results of the capacity test for the models are shown in TTable 4.2. The table contains also the reversible and operational capacity, calculated internally by the software.

Table 44.22: The capacity test results for the two electrochemical models.

NCM111 NCM523

Reversible capacity calculated internally by the software (Ah)

21.769 22.707

Operational capacity calculated internally by the software (Ah)

@4.2V (OCV@100%SOC)

21.558

@4.08329V (OCV@100%SOC)

22.707

Discharge capacity test (Ah) 20.34 20.02

SOC of the model when the voltage reaches to the low cutoff voltage (%)

5.63 11.82

For the electrochemical models, similar to the ECM and the MATLAB model, the calculated SOC by the software, did not show 0%SOC when the voltage reached the low cut off voltage. The reason is, the definition of the capacity in the software is different from the definition, which was presented for the physical tests.

For the model NCM111, 20A discharge current was applied to the model and the SOC of the cell when the voltages reached to the low cut off voltage was 5.63%SOC, this value became around 1%SOC when 0.2A discharge current was applied to the model. It means the calculated capacity by the software is for the infinity low C-rate.

4.1.3 Analysis

Table 4.3 summarizes all the results of capacity tests for the five references and the three designed models.

67


TTable 44..33:: The results of all the capacity tests.

References Models

CSDS cell id 32 cell id 33 cell id 35 Research report

MATLAB NCM111 NCM523

Capacity test (Ah) 20.20 19.71 20.39 20.15 18.3 17.89 20.34 20.02

SOC (%) at low cut off voltage - - - 01.94 - 14.79 5.63 11.82

Calculated capacity (Ah) 20.20 19.71 20.39 20.54 18.3 20.53 21.48 22.38

In TTable 4.3, the SOC at low cut off voltage converted to capacity and added to the value of the capacity test and, it is called calculated capacity. For better understanding, the calculated capacity in Table 4.3 is depicted as a histogram and is shown in Figure 4.4.

Figure 44.44: The histogram of the calculated capacity.

By neglecting the research report, the other four references show 0.68Ah difference from each other. The calculated capacity of the both electrochemical models is around 1Ah higher than the highest capacity in the references.

These differences are due to two main reasons. Firstly, the definition of capacity for the physical tests and electrochemical models are not the same and it should be taken into consideration during the designing process of the electrochemical models. Secondly, the method for calculating the capacity of the physical cells and the designed models are not exactly the same in this thesis because of some technical problems which mentioned before.

20.2 19.71 20.39 20.5418.3

20.53 21.48 22.38

0

5

10

15

20

25

CSDS Cell id 32 Cell id 33 Cell id 35 Researchreport

MATLABmodel

NCM111 NCM523

Calculated capacity (Ah)

68


4.2 Voltage during the 1C discharge current test

This test represents the voltage of the cell during a constant 1C discharge current. This test is performed at room temperature, like all the other tests. Moreover, due to different capacities for the references/models and for a better comparison all the graphs for this test are drawn in voltages versus SOC instead of time or capacity.

4.2.1 References

In this test, the models are compared with three references, the physical test on the cell id 32, ECM and CSDS. The physical test on the cell id 33 is not considered in this test as reference, because, although the profile of these two physical tests are similar, they are not exactly the same and then they cannot be compared together. Therefore the cell id 32 is selected, because the physical/chemical measurements during the design process were done on this cell.

4.2.1.1 CSDS

Based on the statistical study showed in Appendix A, in 93% of the studied CSDS, the discharge graph is available for different C-rates. This graph from the CSDS can be used as a good reference for the capacity test. FFigure 4.2 shows the mentioned graph called “rate capability” in CSDS of the sample cell. The selected C-rate for this test is 1C-rate, the light green curve. By reading 22 points of the curve by naked eye and also by considering 20.2Ah as fully SOC, the Voltage-SOC curve can be drawn as it shown in Figure 4.5.

Figure 44.55: Voltage-SOC curve of the CSDS during the 1C discharge current.

69


The curve is not smooth like the original curve in FFigure 4.2 and the reason is the data points were read by the naked eye. This can cause some negligible errors.

4.2.1.2 Physical test on the cell id 32

The formation test for the cell id 32, showed in Figure 4.1, is used here for drawing the voltage response of the cell during the 1C continuous current. The calculated capacity for this cell is 19.71Ah. Then the applied current (1C-rate) should be 19.71A, but there was not any physical test with this applied current and inevitably, the same test results with current 20A, were used.

For drawing the Voltage-SOC, the output voltage showed in Figure 4.1(a), is used, but the SOC% of the cell should be calculated from the current and the time. For each measured point, the dropped Ah calculated by multiplying the amplitude of the current and the duration of the sample in hour. The results of the calculations for SOC and the overall voltage-SOC curve are drawn in Figure 4.6.

Figure 44.66: Voltage-SOC curve of the physical test on the cell id 32 during the 1C discharge current.

The voltage of the cell, showed in Figure 4.6, dropped to below 4V after applying the discharge pulse and it can be due to the high contact resistance. Probably the current collectors of the cell did not weld to the cables and this cause such a high contact resistance and, consequently, a high dropped voltage.

70



In this test, the calculated capacity of the cell id 35, was considered for applying the 1C current. Then 20.54A, as 1C discharge current, was applied to the model for one hour. The result is shown in FFigure 4.7. The figure shows the voltage of the cell versus the SOC percentage of the model.

Figure 44.77: Voltage-SOC curve of the ECM-cell id 35 during the 1C discharge current.

After applying one hour 1-C discharge current, the SOC% of the model reached the value of 0.029%SOC, which can be considered as zero. Also this proves, the capacity 20.54Ah is calculated correctly for the ECM.

4.2.2 Models

For this test, four models are available, three designed models and an extra model, called “electrochemical NCM111*”. This model is the same as NCM111, but with a different value for the parameter OCV@100%SOC. The reason of this change needs to be explained in detail. In electrochemical models, three parameters: OCV@100%SOC and contact resistances of electrodes (anode and cathode) have a considerable effect to the models behavior.

Two parameters, contact resistance of anode and cathode (explained in section 3.2.2.4), represent the contact resistance between the positive/negative electrodes and the foil current collectors. The value of these parameters were selected in the designed models

71


somehow to being close enough to the ECM (cell id 35). The reason for this decision was that the contact resistance in physical tests on the cell id 35 is more reasonable than the cell ids 32 and 33.

The parameter OCV@100%SOC is one of the parameters in the initial conditions tab. This parameter is considered as key parameter and explained in section 3.2.2.3. The value of this parameter for the electrochemical model NCM111 is 4.2V and it is possible to change it to the lower values too. But the situation for the electrochemical model NCM523 is different. This model refuses the values higher than 4.075 for OCV@100%SOC. The error message from the software is “Specified voltage@100%SOC=4.2V is too high. This might be caused by cathode limited formation”. In this situation and for understanding the effect of this value to the results, the model NCM111* is defined and added to the designed models. The value of 4.0V is selected for the NCM111*.


The calculated capacity for the MATLAB generic battery model is 20.53Ah, then a 20.53A current was applied to the model for one hour. The output voltage and SOC of the models are used for drawing the voltage-SOC, shown in FFigure 4.8.

Figure 44.88: Voltage-SOC curve of the MATLAB model during the 1C discharge current.

As there is not any option for cutoff voltage in the MATLAB model, the model continues to reaches 0V and stay on this voltage until the SOC reaches 0%. The results below of 2.5V is not reliable.

72



Based on the calculated capacities, currents 21.48A and 22.38A were applied to the models NCM111 and respectively, NCM523 for one hour. For the model NCM111* the same input as for the NCM111 was used. The Voltage-SOC curves are drawn in FFigure 4.9.

Figure 44.99: Voltage-SOC curves of the electrochemical models during the 1C discharge current.

The first interesting point in Figure 4.9 is, none of the electrochemical models could reach 0%SOC prior to the low cutoff voltage. The SOCs of the models when the voltage reached the low cutoff voltage were 12.09%SOC, 7.2%SOC and 5.9%SOC for the models NCM523, NCM111* and, respectively, NCM111. The reason as mentioned in section 4.1.2.2, is the difference of capacity definition in the software. It is worth to mention that the SOC of the electrochemical model was a line during the whole test and it means there was not any correction of SOC inside of the software.

Another interesting point in Figure 4.9 is the behavior of the model NCM111*. Two models have a difference of 0.2V at 100%SOC, but their difference is just 0.019V at 15%SOC.

In the overall, the three designed models at 50%SOC have difference voltage around 0.06V. This can be due to different capacities and, also, the effect of the parameters: OCV@100%SOC and contact resistances, which should be selected precisely.

73


4.2.3 Analysis

As a summary for the voltage test, all of the results for the three references and the four models are compared together. Seven curves are drawn in FFigure 4.10 for the all voltage tests.

Figure 44.110: Voltage-SOC curves of all models and references (full SOC range).

Between all of the curves in Figure 4.10, the MATLAB generic battery model has a different shape and far away from the references. For a better comparison, Figure 4.11 with a zoom in SOC range 80% till 30% is drawn.

Figure 44.111: Voltage-SOC curves of all models and references (SOC range 80%-30%).

74


In FFigure 4.11, the cell id 32 has the lowest voltage and by neglecting the MATLAB model, the electrochemical NCM111 model has the highest voltage for each SOC%.

The model NCM111 has an offset around 0.045V-0.065V from the CSDS curve. The NCM111* could fill this offset and this shows the effect of OCV@100%SOC parameter to the model’s results. The behavior of the model NCM523 is similar to the model NCM111*.

Giving a number as the accuracy of the designed models is avoided here due to the fact that there are differences between the references, too. In the overall, the model NCM523 could show an acceptable result. Also, the electrochemical NCM111 could show a good result if the mentioned effective parameters are select correctly. This was tested with the electrochemical model NCM111*.

4.3 Open circuit voltage test

OCV is the voltage measured on a battery when it is disconnected from an electrical circuit and not under any load. This test is performed to gain information about the OCV after charge and discharge. If a cell cannot show a good accuracy in the OCV test, the results of the other tests like the HPPC test cannot be accurate either. Then before the HPPC test, the OCV test is presented.

For acquiring the OCV-SOC curve, in simulation tests, a 0A current was applied to the models and the output voltage for each initial SOC was considered as OCV, but in physical tests, the cell is charging and discharging to a certain SOC and the voltages after the rest time measured as OCV. The test data was analyzed by firstly, computing a voltage relationship versus SOC for the discharging portion of the test, and then computing a second voltage relationship versus SOC for the charging portion of the test.

4.3.1 References

For this test, the designed models are compared with three references: physical test on the cell id 33, the ECM (cell id 35) and the research report.

4.3.1.1 Physical test on the cell id 33

As mentioned before, for acquiring the OCV-SOC curve in physical tests, the cell should be charged and discharged to different levels of SOC, then at a time after the charging or discharging has stopped and the battery has reached an equilibrium settled state, the voltage is measured as OCV at each of the different levels of SOC. Between the all done tests on the cell ids 32 and 33, the lowest amplitude is 15A for discharge and 5A for charge. This test is called “dynamic resistance and open circuit voltage”. The test

75


contains one charge and discharge cycle and it takes more than 21 hours. The input currents and the output voltages of this test are shown in FFigure 4.12.

(a)

(b)

Figure 44.112: The dynamic resistance and open circuit voltage test. (a): The applied current pulses to the physical cell. (b): The voltage response from the physical cell.

The “dynamic resistance and open circuit voltage” test contains 18 discharge and 47 charge pulses. The voltages at the end of the rest times between the pulses were taken as OCV (the points showed by the green arrows in Figure 4.12(b)).

76


The amplitude of the current pulses as it can be seen in FFigure 4.12(a) is 15A, around 0.73C-rate for the discharge part and 5A, around 0.24C-rate for the charge part. Each pulse is followed by a rest period of at least 15minutes to allow the battery to reach a semi-equilibrium state. In theory, this rest period must be extended to a several hours to be able to measure the true OCV, but in practice a shorter time is used to reduce the total test duration.

The voltage of the cell starts from 4.2V, which means fully charged and returns again to 4.2V (more accurately to 4.202V) which again can be considered as fully charged. By calculating the dropped SOC from the Figure 4.12(a) for each sample point and finding the corresponded voltage in Figure 4.12(b) the OCV for each level of SOC is achievable. This can be done for the charge part and as well as for the discharge part. The procedure is similar to calculating the OCV for the voltage test showed in section 4.2.1.2.

The calculated OCV and also extracted OCV-SOC curves for the charge and discharge parts are shown in Figure 4.13.

Figure 44.113: OCV calculations for the physical test on the cell id 33.

The blue pulses in Figure 4.13 show the mentioned calculations for each measured point. For extracting the OCV-SOC curves for charging and discharging parts, the voltages at the end of the rest time should be selected, as it is shown in Figure 4.12(b). In Figure 4.13, the selected OCV points are 18 points for discharging part, showed with green “×” marks and 47 points for charging part, showed with red “*” marks.

77


The calculated voltages for the discharge part are lower and for the charge part are higher than the real voltages, because the rest time is just 15 minutes instead of at least 1-2 hours.


For the physical cell id 35, instead of the physical test results, the look up table of OCV-SOC of the model was used. The lookup table is based on the “dynamic resistance and open circuit voltage” test, which was explained and showed in section 4.3.1.1. For the case of selecting the room temperature 23 , the OCV-SOC curve of the cell id 35 is drawn in FFigure 4.14.

Figure 44.114: OCV- SOC curve of the ECM-cell id 35.

4.3.1.3 Research report

In the research report [37], in addition of the capacity measurements, which was shown in section 4.1.1.4, the OCV-SOC curve is also estimated. Siqi Lin in her report explained that the OCV-SOC curves are estimated by charging and discharging the cell with rate. The charge and discharge curves from the research report are shown in Figure 4.15. OCV-CH curve is the result of the subtraction of CH_C 10 from , polarization voltage. Also the OCV-DCH curve is the result of the sum of DCH_C 10 and . The polarization voltage which is representing the change of cell voltage from its open-circuit voltage, will appear during the charge and discharge state. For understanding the polarization voltage and its effect, the readers are referred to [38].

78


FFigure 44..115:: OCV-SOC curves from the research report [37].

In FFigure 4.15, the label of X-axis is SOC (%). In this graph, the effect of hysteresis is around 0.04V which is a normal value for NCM chemistry. Both OCV-DCH and OCV-CH curves will be used as references in section 4.3.3.

4.3.2 Models

For the OCV test, there are three models: MATLAB generic battery model and electrochemical models NCM111 and NCM523. Although testing the electrochemical NCM111* could be interesting, performing the OCV test was not possible for this one, due to the expired license of the software. The explanation of the electrochemical NCM111* was given in section 4.2.2.


For acquiring the OCV-SOC curve of the MATLAB model, the initial SOC of the model was changed for each 1%SOC in a loop and by applying 0A current to the model, the output voltage was considered as OCV. The OCV-SOC curve is depicted in Figure 4.16. This curve contains 101 points.

79


FFigure 44..116:: OCV-SOC curve of the MATLAB model.


For acquiring the OCV-SOC for the electrochemical models, the same procedure of the MATLAB model was repeated for these models, too. Due to the structure of the software, less SOC levels were used. Also in the electrochemical model, one of the output ports is OCV, which shows the same or very close value of the output voltage port when the input is zero. The results of the OCV test for the both electrochemical models are shown in FFigure 4.17.

Figure 44.117: OCV-SOC curves of the two electrochemical models.

80


There is an offset between the two models. Except of the offset voltage, the curves have similar shape between 20%SOC and 100%SOC.

4.3.3 Analysis

The results of all the three references and the three models are shown in FFigure 4.18. It contains 8 curves, 5 for the references and 3 for the models. The references are the physical test on the cell id 33 (2 curves, OCV charge and OCV discharge), the ECM and the research report (2 curves, OCV charge and OCV discharge).

Figure 44.118: SOC-OCV curves of all models and references (full SOC range).

At the first sight, it can be seen that the MATLAB model could not follow the references, not in value and not in shape either. Except of the MATLAB model, the rest of the curves have a similar shape and all of them are close to each other. Also the electrochemical model NCM523 shows an offset around 0.05V to 0.1V from the rest curves.

For better comparison between the curves, the SOC range 30% till 80% is selected and drawn for the whole results except of the MATLAB model, in Figure 4.19.

81


FFigure 44..119:: SOC-OCV curves of all the models and references except of the

MATLAB model (SOC range 30%-80%).

In FFigure 4.19, the difference between the curves can be seen more clearly. All of them are in a range around 0.1V. This range becomes just 0.05V by neglecting the electrochemical model NCM523. The electrochemical model NCM111 is showed by cyan dash line and it can follow this narrow 0.05V range, especially for the middle SOC, it is very close to the ECM. Then as conclusion, between the designed models, the electrochemical model NCM111 could show good results. Giving a number as accuracy is avoided for this test, too, because there is a difference between the references.

4.4 HPPC test

Hybrid pulse power characterization (HPPC) test is intended to determine dynamic power capability over the device’s useable voltage range. For a detailed description about the HPPC test, the readers are referred to [39].

Between all the represented tests in this thesis, the HPPC test is the toughest test, which can show the dynamic behavior of the designed models. A pulse with high amplitude and short duration is a challenge for the models. Between the available physical tests, there was just one with high amplitude and short duration, available for the cell id 32.

The HPPC test profile of the cell id 32 is shown in Figure 4.20(a). This profile is also used for the other models. Again all the tests are done at the room temperature. The HPPC test profile started with a discharge pulse of 200A amplitude for 11 seconds and it was

82


followed by 3 minutes rest time and, then, a current of 10A discharge current was applied for around 8.5 minutes. Finally, there is a rest time of approximately 30 minutes prior to the next pulse. This process drops the SOC of the cell from 100% to 90% and it was repeated 10 times, and, therefore, the SOC of the cell would reach the 0%SOC at the end of the discharging pulses for a 20Ah LIB.

Similarly, the charge part started with a charge pulse of 20A amplitude for 10 seconds and it was followed by 3 minutes rest time and, then, the current of 10A amplitude was applied for around 12 minutes. Again, there is a rest time of approximately 30 minutes prior to the next pulse. As in the case of the discharging part, this process increased the SOC of the cell by 10% and, therefore, after 10 repetitions of this process, the cell will be fully charged. The output voltage of the cell id 32 at the first of the test is 4.191V, assumed fully charged and at the end of the test returns to 4.199V. The calculated SOC of the cell during the test is shown in FFigure 4.20(b).

(a)

(b)

Figure 44.220: The HPPC test profile. (a): Applied current. (b): Calculated SOC.

83


The SOC in FFigure 4.20(b) reaches to -1.279% in the middle, between the discharge and charge parts. This can be consider as an error due to mistakes done during the physical measurements or capacity calculations.

The structure of the HPPC test is different from the other presented tests and it is separated in two parts, discharge and charge. In each part, all the references and models are presented together.

4.4.1 HPPC discharge test

For the discharge part, three designed models are compared with just one reference, the cell id 32. The ECM could not run this test and the error message “Invalid current range” was returned. By applying the discharge part of the HPPC current profile to the models, the voltage responses are shown in Figure 4.21.

Figure 44.221: All results of the HPPC discharge test (All 10 pulses).

In this case it can be seen that the MATLAB model has unacceptable results for this test. Also, the electrochemical models NCM523 and NCM111 reached the low cutoff voltage in the sixth and the seventh pulses and, then, the simulation stopped for the rest of the pulses.

There are two problems for comparing the models and references in the way showed in Figure 4.21. Firstly, the errors are cumulated during the test and, secondly, the models and references have different capacities, then the applied pulses to the models have different SOCs and these cause an incomparable situation.

84


For solving these problems, one pulse was selected and applied to the models with the same SOC as reference. The fourth pulse with around 70%SOC (showed with green arrows in FFigure 4.20) was selected. As the MATLAB model could not show acceptable results, this model is not involved into studying the forth pulse. The results of the reference and the two electrochemical models are shown in Figure 4.22

Figure 44.222: All results of the HPPC discharge test (Fourth pulse).

Despite, the electrochemical model NCM111 and the reference, have just 0.002V difference voltage before applying the pulse at second 0, but this difference becomes 0.279V at second 11. This difference for electrochemical model NCM523 is 0.1V at second 0 and 0.202V at second 11. At the end of the rest time the difference of the model’s voltages from the reference are 0.019V and 0.093V for the electrochemical models NCM111 and, respectively, NCM523. It means the dropped voltages after this pulse are 0.048V, 0.027V and 0.022V for the reference and electrochemical models NCM111 and NCM523 respectively.

As there is just one reference in this test, then a number as error can be defined and used to evaluate the models. Comparing the internal resistance can show the error of the models from reference. For calculating the internal resistance for a 10 seconds pulse, the formula (4.1) is used.

10 = ( 10 0 ) (4.1)

Calculating the resistance for 10 seconds is typical for HPPC tests. The black arrow shows the time differences between the second 0 and second 10. The applied current in the first

85


second of the pulse is not constant and changes from 0A to -200A, but due to its negligible effects to the results, the amplitude of the current was considered as 200 during the whole 10 seconds.

The calculated internal resistances based on the above formula for the reference and the two models are shown in TTable 4.4.

Table 44.44: Calculated internal resistance (R10s), HPPC discharge test.

Calculated R10s (Ω) Error (%)

Reference-cell id 32 0.0051 0

Electrochemical NCM111 0.0037 27.45


According to the Table 4.4, the error of the designed models are around 27% and 29% for the electrochemical models NCM111 and NCM523 for such a tough HPPC test. It is worth to mention, for the electrochemical models, a fixed-step size (1 second) was used but for the reference a variable-step size (around 0.034 second for the critical points in the HPPC test) was used.

4.4.2 HPPC charge test

Unlike the discharge part, the ECM can run this test and therefore there are two references for this test. By applying the charge part of the HPPC current profile, showed in Figure 4.20(a) to the models, the voltage responses are shown in Figure 4.23.

Figure 44.223: All results of the HPPC charge test (All 10 pulses).

86


For a better view of the results, below there is the previous figure, magnified from 40%SOC to 70%SOC on the four pulses.

FFigure 44..224:: All results of the HPPC charge test (4 middle pulses).

The first interesting point in FFigure 4.24 is that the two references do not match each other. The voltage of the ECM in the sixth pulse is closer to the electrochemical model NCM111 than the physical test. In the other words, the error of the electrochemical model NCM111 from the physical test is less than twice of the error of the ECM. The electrochemical model NCM523 has the highest error from the physical test.

Like the HPPC discharge test, the sixth pulse with around 50%SOC (showed with orange arrows in Figure 4.20) was selected. Also similar to the HPPC discharge test, the SOC of the electrochemical models and the ECM are set up to 50%SOC to have a comparable situation. Again, the MATLAB model will not be involved for studying the sixth pulse. The results are shown in Figure 4.25.

87


FFigure 44..225:: All results of the HPPC charge test (Sixth pulse).

The first interesting result by comparing the FFigure 4.24 and Figure 4.25 is that the ECM and the electrochemical model NCM111 become much closer to each other after correcting the SOCs of them to 50%SOC. Two electrochemical models NCM111 and NCM523 have similar behavior, but with a voltage offset, which probably was caused by the parameter OCV@100%SOC.

Similar to the HPPC discharge test, the error of the designed models is calculated from the two references. The calculated internal resistances, based on the formula (4.1), are shown in Table 4.5.

Table 44.55: Calculated internal resistance (R10s) from the HPPC charge test.

Calculated R10s (Ω) Error (%) from cell id 32

Reference-cell id 32 0.00555 0

ECM (cell id 35) 0.00485 12.6126



The resistance of the ECM (cell id 35) has around 12% error comparing to the reference cell id 32. This is a considerable error and shows that, maybe, the two cells are not from the same manufacture line or they have different versions of NCM chemistry in their cathodes. It can be also due to some differences in the performed physical tests on the cell ids 35 and 32.

88


The resistance error of the designed models are the same for both of them and is around 31% comparing to the cell id 32. This is a considerable error and more than the goal of the thesis which was 15%. In the next chapter, the quality of the models will be discussed.

89

Chapter5. Closure

This chapter presents conclusions as well as an introduction to potential future works.

5.1 Conclusions

In this thesis, three models were created (one MATLAB model and two electrochemical models). According to the results in chapter 4, the MATLAB model did not show acceptable results for our sample cell, which has NCM chemistry in its cathode. In the examples that are shown in the website of Mathworks, for this model, the voltage of the cell in their discharge curves is almost constant between 90%SOC and 30%SOC, but in our case it is not. This can be one of the reasons, why this model could not show acceptable results.

Two electrochemical models were produced by the proposed method. Electrochemical models, due to their complexity and requirements of advanced instruments, are sometimes called “inaccessible” for those who are not cell designers or cell manufacturers. In this thesis, this model was created without having any background in chemistry, which constitutes an advantage of the gray box method.

In section 1.2, the goal of the thesis was presented in the form of the below 5 requirements:

1. Able to produce a generic battery model for different types of the LIB chemistries 2. No or low cost for the development of the model 3. A time span around one week for obtaining the model

55 CClosure

90

Chapter5. Closure

4. Able to predict the most aspects of the battery’s behavior like the voltage, SOC, temperature and, preferably, simulate the degradation effects, safety and thermal aspects

5. Accuracy with less than 15% error

The first three objectives were fully achieved and the fourth and fifth ones just partially. The models were produced with nearly no cost and in a time efficient way (around a week) for a sample cell with NCM chemistry in its cathode. The other chemistries and geometries can be checked with the same method as part of future works. The focus of this thesis was on just predicting the voltage and SOC of the sample cell. As part of future works, the other capabilities, showed in TTable 5.1, can also be examined.

Table 55.11: Comparison of the black box and the gray box methods.

Black-box method

Gray-box method

without tuning

Gray-box method with tuning (Expected)

Model ECM (without

aging/life simulation)

Electrochemical Electrochemical

Development Time 3 months 1 week 2-3 weeks

Cost

Full physical tests for different

temperatures, current amplitudes

SEM images Material analysis

Formation test HPPC test SEM images Material analysis

Computational Effort Efficient Acceptable Acceptable

Error 1%-5% Around 30% Less than 30%

Cap

abili

ties

Predicting V, SOC Predicting Temperature Life/aging simulation 3D-thermal modeling Safety simulation

The models were tested in chapter 4 to check their accuracy and they showed good results for the OCV and voltage tests. The results in these two tests accompany the references. For the HPPC tests, the calculated 10 seconds resistances were around 30% less than the references. The error in SOC estimation should also be taken into consideration. The overall errors are higher than the goal of the thesis, which was expected to be 15% error.

91

Chapter5. Closure

The main two reasons behind this are using a new NCM chemistry, which imposed more uncertainty in the model, and skipping the parameters’ tuning process due to lack of time. It is expected that this process, which is one of the important modeling steps, can reduce the error of the models considerably.

TTable 5.1 shows a comparison between the black box and the gray box methods with big improvements in cost and time, which were the most important goals of the thesis, but these achievements were acquired by sacrificing the accuracy.

In conclusion, the achieved results are promising for the idea of the gray box that is in its nascent stages and needs time to be developed and to get used for commercial purposes.

5.2 Future Work

As mentioned before, the proposed method is in its first steps and a lot of improvements can be made. In this thesis, the efforts to reveal the potential of sources are summarized as below:

Knowledge-based: SSufficient efforts Physical/chemical measurements: SSome efforts Physical tests: IInsufficient efforts

Then the first task for future works can be revealing the potential of the physical tests. Tuning the parameters can be undertaken as an unfinished work. If the tuning process could reduce the errors as expected, it would lead to a big improvement in accuracy. This tuning process as mentioned in section 3.2.1.3, can be used for identifying of the key parameters.

The second task could be putting more efforts on the second source, the physical/chemical measurements. As the first step, finding all of the key parameters are necessary. The second step is studying the key parameters for creating a more accurate model.

The gray box method can be used to create other models like ECM. This could be the third task for future works.

In conclusion, if the above three tasks are improving the accuracy of the voltage and SOC prediction, then other capabilities like temperature, safety, life/aging simulation and 3D thermal modeling can be also investigated in the future. Other suggestions, regarding the future works, were given in different parts of the thesis.

92

References

References

[1] “All-Electric Vehicles.” [Online]. Available: https://www.fueleconomy.gov/feg/evtech.shtml. [Accessed: 15-Feb-2016].

[2] “UNSD // Global Forum on International Trade Statistics and Economic Globalization.” [Online]. Available: http://unstats.un.org/unsd/trade/globalforum/tsbPublication/pubsPublic/?currentFilterF=2&currentFilterT=US%20value%20chain. [Accessed: 05-Aug-2015].

[3] L. W. Yao and J. A. Aziz, “Modeling of Lithium Ion battery with nonlinear transfer resistance,” in 2011 IEEE Applied Power Electronics Colloquium (IAPEC), 2011, pp. 104–109.

[4] R. V. Kumar and T. Sarakonsri, “Introduction to Electrochemical Cells,” in High Energy Density Lithium Batteries: Materials, Engineering, Applications, Wiley‐VCH Verlag GmbH & Co. KGaA, 2010, pp. 1–25.

[5] “battery news - Global Battery… - Something About… - Comparing the… - Four Renegades of… - Manganese Finds New… - aubattery.over-blog.com’s name.” [Online]. Available: http://aubattery.over-blog.com/categorie-11757620.html. [Accessed: 23-Jun-2015].

[6] “LG Chem Power Inc - High-powered, high-capacity lithium ion batteries.” [Online]. Available: http://www.lgcpi.com/lithium.shtml. [Accessed: 06-Aug-2015].

[7] “Lithium polymer battery,” Wikipedia, the free encyclopedia. 14-Dec-2015. [8] “Assessing The Safety Of Lithium-Ion Batteries | February 11, 2013 Issue - Vol. 91

Issue 6 | Chemical & Engineering News.” [Online]. Available: http://cen.acs.org/articles/91/i6/Assessing-Safety-Lithium-Ion-Batteries.html. [Accessed: 06-Dec-2015].

[9] “Wiley: Battery Systems Engineering - Christopher D. Rahn, Chao-Yang Wang.” [Online]. Available: http://eu.wiley.com/WileyCDA/WileyTitle/productCd-1119979501.html. [Accessed: 04-Jan-2016].

[10] R.-J. Chen, T.-D. Lin, H.-Y. Lin, and Y.-C. Hsieh, “Electric circuit modelling for lithium-ion batteries by intermittent discharging,” IET Power Electron., vol. 7, no. 10, pp. 2672–2677, Oct. 2014.

[11] R. van Yperen, A. . van Dillen, J. . Geus, E. Boellaard, A. . van der Horst, and A. . van der Kraan, “Regenerable sorbent for high-temperature desulfurization based on iron-molybdenum mixed oxides,” Stud. Surf. Sci. Catal., vol. 91, no. C, pp. 579 – 588, 1995.

93

References

[12] M. Chen and G. A. Rincon-Mora, “Accurate Electrical Battery Model Capable of Predicting Runtime and I–V Performance,” IEEE Trans. Energy Convers., vol. 21, no. 2, pp. 504–511, Jun. 2006.

[13] W. Blackstone, T. M. Cooley, and J. D. W. Andrews, “Commentaries on the laws of England: in four books,” 1899. [Online]. Available: http://infotrac.galegroup.com/cookie_cutter?URL=http://0-galenet.galegroup.com.bianca.penlib.du.edu/servlet/MOML?af=RN&ae=F101737552&srchtp=a&ste=14&q=denv80220. [Accessed: 25-Jun-2015].

[14] J. A. Pearsol, “Justifying conclusions in naturalistic evaluations. Introduction to special section,” Eval. Program Plann., vol. 10, no. 4, pp. 307 – 308, 1987.

[15] D. Rakhmatov, S. Vrudhula, and D. A. Wallach, “A model for battery lifetime analysis for organizing applications on a pocket computer,” IEEE Trans. Very Large Scale Integr. VLSI Syst., vol. 11, no. 6, pp. 1019–1030, Dec. 2003.

[16] L. Gao, S. Liu, and R. A. Dougal, “Dynamic lithium-ion battery model for system simulation,” IEEE Trans. Compon. Packag. Technol., vol. 25, no. 3, pp. 495–505, Sep. 2002.

[17] “PNGV Battery Test Manual Revision 3 - pngv_manual_rev3b.pdf.” [Online]. Available: http://avt.inel.gov/battery/pdf/pngv_manual_rev3b.pdf. [Accessed: 06-Dec-2015].

[18] J. Newman and W. Tiedemann, “Porous-electrode theory with battery applications,” AIChE J., vol. 21, no. 1, pp. 25–41, Jan. 1975.

[19] M. Doyle, T. F. Fuller, and J. Newman, “Modeling of Galvanostatic Charge and Discharge of the Lithium/Polymer/Insertion Cell,” J. Electrochem. Soc., vol. 140, no. 6, pp. 1526–1533, Jun. 1993.

[20] V. R. Subramanian, V. Boovaragavan, V. Ramadesigan, and M. Arabandi, “Mathematical Model Reformulation for Lithium-Ion Battery Simulations: Galvanostatic Boundary Conditions,” J. Electrochem. Soc., vol. 156, no. 4, pp. A260–A271, Apr. 2009.

[21] K. M. Tsang, L. Sun, and W. L. Chan, “Identification and modelling of Lithium ion battery,” Energy Convers. Manag., vol. 51, no. 12, pp. 2857–2862, Dec. 2010.

[22] X. Yuan, H. Liu, and J. Zhang, Lithium-Ion Batteries: Advanced Materials and Technologies. Hoboken: CRC Press, 2011.

[23] W. B. Gu and C.-Y. Wang, “Thermal and electrochemical coupled modeling of a lithium-ion cell,” Proc-Electrochem Soc, vol. 99, pp. 748–762, 2000.

[24] V. Srinivasan and C. Y. Wang, “Analysis of electrochemical and thermal behavior of Li-ion cells,” J. Electrochem. Soc., vol. 150, no. 1, pp. A98–A106, 2003.

[25] K. Smith and C.-Y. Wang, “Power and thermal characterization of a lithium-ion battery pack for hybrid-electric vehicles,” J. Power Sources, vol. 160, no. 1, pp. 662–673, 2006.

94

References

[26] P. Taheri and M. Bahrami, “Temperature Rise in Prismatic Polymer Lithium-Ion Batteries: An Analytic Approach,” SAE Int. J. Passeng. Cars - Electron. Electr. Syst., vol. 5, no. 1, pp. 164–176, Apr. 2012.

[27] T. V. S. L. Satyavani, B. Ramya Kiran, V. Rajesh Kumar, A. Srinivas Kumar, and S. V. Naidu, “Effect of particle size on dc conductivity, activation energy and diffusion coefficient of lithium iron phosphate in Li-ion cells,” Eng. Sci. Technol. Int. J.

[28] Y. Zhang, Z.-B. Wang, J. Lei, F.-F. Li, J. Wu, X.-G. Zhang, F.-D. Yu, and K. Ke, “Investigation on performance of Li(Ni0.5Co0.2Mn0.3)1−xTixO2 cathode materials for lithium-ion battery,” Ceram. Int., vol. 41, no. 7, pp. 9069–9077, Aug. 2015.

[29] P. Taheri, S. Hsieh, and M. Bahrami, “Investigating electrical contact resistance losses in lithium-ion battery assemblies for hybrid and electric vehicles,” J. Power Sources, vol. 196, no. 15, pp. 6525–6533, Aug. 2011.

[30] A. A. Andriiko, P. V. Rudenok, and L. I. Nyrkova, “Diffusion coefficient of Li+ in solid-state rechargeable battery materials,” J. Power Sources, vol. 72, no. 2, pp. 146–149, Apr. 1998.

[31] M. Ma and B. C. Sisk, “Investigation of Solid State Diffusion Coefficient Dependency in Lithium Ion Battery,” Meet. Abstr., vol. MA2014–02, no. 5, pp. 404–404, Aug. 2014.

[32] K. Smith and C.-Y. Wang, “Solid-state diffusion limitations on pulse operation of a lithium ion cell for hybrid electric vehicles,” J. Power Sources, vol. 161, no. 1, pp. 628–639, Oct. 2006.

[33] D.-W. Chung, M. Ebner, D. R. Ely, V. Wood, and R. E. García, “Validity of the Bruggeman relation for porous electrodes,” Model. Simul. Mater. Sci. Eng., vol. 21, no. 7, p. 074009, Oct. 2013.

[34] A. Vadakkepatt, B. Trembacki, S. R. Mathur, and J. Y. Murthy, “Bruggeman’s Exponents for Effective Thermal Conductivity of Lithium-Ion Battery Electrodes,” J. Electrochem. Soc., vol. 163, no. 2, pp. A119–A130, Jan. 2016.

[35] K. K. Patel, J. M. Paulsen, and J. Desilvestro, “Numerical simulation of porous networks in relation to battery electrodes and separators,” J. Power Sources, vol. 122, no. 2, pp. 144–152, Jul. 2003.

[36] M. Ebner and V. Wood, “Tool for Tortuosity Estimation in Lithium Ion Battery Porous Electrodes,” J. Electrochem. Soc., vol. 162, no. 2, pp. A3064–A3070, Jan. 2015.

[37] “Publication Detail – Institute of Transportation Studies – Davis.” [Online]. Available: http://www.its.ucdavis.edu/research/publications/publication-detail/?pub_id=1720. [Accessed: 04-Jan-2016].

95

References

[38] Y. Jiang, C. Zhang, W. Zhang, W. Shi, and Q. Liu, “Modeling charge polarization voltage for large lithium-ion batteries in electric vehicles,” J. Ind. Eng. Manag., vol. 6, no. 2, Jun. 2013.

[39] “Microsoft Word - PLUG_IN_HYBRID_Manual Rev 2.doc - PLUG_IN_HYBRID_Manual Rev 2.pdf.” [Online]. Available: http://avt.inl.gov/battery/pdf/PLUG_IN_HYBRID_Manual%20Rev%202.pdf. [Accessed: 01-Dec-2015].

[40] “NREL: Energy Storage - Computer-Aided Engineering for Electric-Drive Vehicle Batteries (CAEBAT).” [Online]. Available: http://www.nrel.gov/transportation/energystorage/caebat.html. [Accessed: 07-Aug-2015].

96

Statistical study of information in CSDS & MSDS

The total number of studied cells in the statistical study are 15. The selected cells for study are from the most well-known brands like Samsung, LG, Sony, Panasonic, Boston-Power, A123, E-One Moli Energy and EIG. Eight of these cells are energy optimized, six are power optimized and for one is unknown. More over the number of cells with cylindrical geometry are 8 and with prismatic geometry (stacked and rolled) are 7. The chemistries of the studied cells are shown in FFigure A.1.

Figure AA.11: Chemistry of the studied cells in the statistical study.

The results of the statistical study are shown in Table A.1.

LiCoO226.67%

LiNiO220.00%LFP 26.67%

NMC 20.00%

NCA 6.67%

Appendix A. SStatistical study of information in CSDS & MSDS

97

Statistical study of information in CSDS & MSDS

TTable AA..11:: The results of the statistical study.

# Information Availability (%)

1 Nominal Capacity (Ah) 100 2 Nominal Voltage (V) 100 3 Geometry 100 4 Dimensions 100 5 Weight 100 6 Year of production 93 7 Specific Power (W/kg) 40 8 Specific Energy (Wh/kg) 67 9 Power density (W/L) 40

10 Energy Density (Wh/L) 53 11 Energy content or Nominal Energy (Wh) 53 12 Cathode material 100 13 Approximate weight percentage of Cathode 80 14 Anode material 93 15 Approximate weight percentage of Anode 60 16 Electrolyte material 47 17 Approximate weight percentage of Electrolyte 67 18 Recommended Standard Charge/Discharge Method 67 19 Recommended Fast Charge Method to 80% SOC 27 20 Nominal discharge Power (W) 13 21 Maximum continuous discharge (A) 40 22 Maximum Pulse Discharge (A) (at 10 sec) 13 23 HPPC* (10sec Discharge pulse power 50% SOC) 27 24 Discharge Cut off Voltage (V) 33 25 Graph-Discharge power capability (Watts) vs SOC (%) ( 13 26 Graph-Voltage(V) during discharge - varying Watts 13

27 Graph-Voltage(V) during discharge (capacity Ah) (at fixed C-rate) - varying temperature 33

28 Graph-Voltage (V) during discharge -varying C-rate 80 29 Graph-Voltage (V) during charge/discharge , fixed current) 87 30 Graph-Current (mA) during charge time (room temp, fixed current) 67 31 Internal impedance 73 32 Internal resistance (DC) 13 33 Operation and or storage Temperature 93 34 Graph-Cycle life 20 35 Graph-Capacity (% or Ah) vs Cycles (numbers) (at % DOD) 53 36 Graph-Capacity degradation (%) vs Energy 7 37 Graph-Capacity (%) vs Temperature C) varying C-rate 47

98

CAEBAT

According to [40], computer-aided engineering for electric-drive vehicle batteries (CAEBAT), is a project for accelerating the development and lowering the cost of lithium-ion batteries for next-generation electric-drive vehicles (EDVs). CAEBAT efforts focus on:

Developing engineering tools to design battery cells and packs Shortening the battery prototyping and manufacturing processes Improving overall battery performance, safety, and lifespan Reducing expenses related to battery development and production

The vehicle technologies office (VTO) of the U.S. department of energy's office of energy efficiency and renewable energy initiated the CAEBAT project in 2010 to facilitate the development of computer-aided engineering tools based on battery models developed by the national laboratories.

NREL brings its predictive computer simulation of LIBs, known as a multi-scale multi-dimensional (MSMD) model framework, to the CAEBAT project. MSMD's modular, flexible architecture connects the physics of battery charge/discharge processes, thermal control, safety, and reliability in a computationally efficient manner. This allows independent development of submodels at the cell and pack levels.

NREL coordinates CAEBAT activities with battery, vehicle, and software industries to develop the first generation of electrochemical-thermal CAEBAT tools for the design and simulation of battery cells and packs.

After a competitive selection process, NREL awarded subcontracts worth $7 million to the following three industry teams:

EC Power (AutoLion software), Penn State University, Johnson Controls, Inc., and Ford

AAppendix B. CCAEBAT

99

CAEBAT

General Motors, ANSYS, and ESim CD-adapco (Battery Design LLC), A123 Systems, and Johnson Controls Inc.

Each CAEBAT team is working independently to develop and validate modeling and design tools for EDV batteries, with an emphasis on integrating electrochemical, electrical, mechanical, and thermal physics. Teams are also exploring different chemistries, cell geometries, and battery pack configurations.

In support of the CAEBAT project, Oak Ridge national laboratory (ORNL) is developing an open-architecture software interface to link the models developed by different teams into the CAEBAT suite of tools. ORNL is also developing input-output interfaces to allow utilization of models across different platforms.

For more information the readers are referred to the website of this project at http://www.nrel.gov/transportation/energystorage/caebat.html and the website of the software at http://batterysim.org.

100

Safety instruction for opening the cell

Opening a physical battery cell is dangerous, if the safety instruction do not followed correctly. The electrolyte inside of the battery is extremely flammable and there is a high risk of fire.

This safety instruction was written by the help of Mr. Kjell Johansson from ÅF Automotive. The MSDS of the cell contains the specialized information regarding the safety instructions and then should be read and considered before opening the cell.

Hazard class LIB hazard is considered in the class 9, miscellaneous. The miscellaneous hazardous material is a material that presents a hazard during transportation but which does not meet the definition of any other hazard class. Miscellaneous dangerous goods present a wide array of potential hazards to human health and safety, infrastructure and or their means of transport. FFigure C.1 shows the name and symbols of all 9 classes of hazardous materials, taken from the website of the U.S. department of transportation.

Appendix C. SSafety instruction for opening the cell

101


FFigure CC..11:: Nine classes of hazardous materials.

During opening the cell The below safety instructions before and during the opening process should be considered:

The first step for opening the cell is, fully discharging the cell by an external load and securing of the fully discharged by measuring the voltage In the whole process before and after opening the cell, short circuit should be avoided seriously Using the protective glasses and gloves is mandatory for the person who wants to open the cell during the whole process Using the metal cutters should be prevented and plastic or insulated cutters should be used instead. FFigure C.2 shows a simple plastic cutter made by the laboratory personnel

Figure CC.22: The plastic cutter was used for opening the cell.

During opening the cell, hydrogen fluoride and carbon monoxide may be release, then a ventilation cabin is necessary for preventing of toxic gases inhalation. Figure C.3 shows a ventilation cabin which is suitable during the opening and measurement process

102


FFigure CC..33:: The ventilation cabins.

During the opening process, small amounts of electrolyte may leak. Sand or earth should be used to absorb any exuded material For putting the cell inside of the ventilation cabin or on a table, the surface should be electrically insulated and heat resistant The insulation surface should be used for the measurement instruments. FFigure C.4 shows the cell on the balance for measuring the weight of the cell. A plastic insulator was used on the metallic surface of the balance

Figure CC.44: A plastic insulator was used for measuring the weight of the cell with

the balance.

The battery cell should not be expose to extreme heat or open flames and try to keep the battery at room temperature and preferably in cool and dry place

103


In case of fire, a self-contained breathing apparatus (SCBA) and an extinguishing media, suitable for the material, are required In case of fire, the cell should be removed from the fire-fighting area immediately due to high explode risk. The cell may explode if heated above 125 If the skin has come into contact with the electrolyte, it should be washed thoroughly with water

First aid During and after the opening process and in the case of below events, the following actions are recommended:

IInhalation: Leave area immediately and seek medical attention Eye contact: Rinse eyes with water for 15 minutes and seek medical attention Skin contact: Wash area thoroughly with soap and water and seek medical attention Ingestion: Drink milk/water and induce vomiting, seek medical attention

For medical assistance in Sweden and as well as the whole European counties, call the free number 112.

After finishing the measurement The cell should not be dump into any sewers, on the ground or into any body of water. Please dispose of or recycle in accordance with appropriate local regulations The cell should be placed in a sealed plastic bag or container and disposed of according to local regulations

104

LLast page

Last page

Last page

105

Documents

Prediction and analysis of model’s parameters of Li-ion ...918015/FULLTEXT01.pdf · Prediction and analysis of model’s parameters of Li-ion battery cells Ali Dareini ... Abbreviations